From 7ed017559b74354b318f2197fb2ca742a9439e7b Mon Sep 17 00:00:00 2001 From: zhurui <274461951@qq.com> Date: Mon, 17 Jun 2024 14:04:28 +0800 Subject: [PATCH] commit --- .gitignore | 4 + README.md | 0 main.py | 176 ++ method3_dict.txt | 20 + result.json | 1 + result1.json | 518 ++++++ result_arxiv_knowledge_graph.json | 1 + t1.py | 14 + te_u/arxiv.py | 150 ++ te_u/paper_down_load/csv/ECCV_2022.csv | 1646 +++++++++++++++++ te_u/paper_down_load/eccv_download.py | 658 +++++++ te_u/paper_down_load/pdf_show.py | 9 + te_u/paper_down_load/pdf_show2.py | 64 + .../urls/init_url_ECCV_2022.dat | Bin 0 -> 2299323 bytes te_u/result_arxiv_knowledge_graph.json | 32 + temp.py | 2 + test_textrank_en.py | 160 ++ test_textrank_zh.py | 18 + utils.py | 157 ++ 小实验/t.json | 168 ++ 小实验/网页访问gpt-4.py | 129 ++ 小实验/网页访问gpt-4——上传文件.py | 137 ++ ...息爬取（题目、期刊、日期、摘要、关键词）_1.py | 282 +++ 23 files changed, 4346 insertions(+) create mode 100644 .gitignore create mode 100644 README.md create mode 100644 main.py create mode 100644 method3_dict.txt create mode 100644 result.json create mode 100644 result1.json create mode 100644 result_arxiv_knowledge_graph.json create mode 100644 t1.py create mode 100644 te_u/arxiv.py create mode 100644 te_u/paper_down_load/csv/ECCV_2022.csv create mode 100644 te_u/paper_down_load/eccv_download.py create mode 100644 te_u/paper_down_load/pdf_show.py create mode 100644 te_u/paper_down_load/pdf_show2.py create mode 100644 te_u/paper_down_load/urls/init_url_ECCV_2022.dat create mode 100644 te_u/result_arxiv_knowledge_graph.json create mode 100644 temp.py create mode 100644 test_textrank_en.py create mode 100644 test_textrank_zh.py create mode 100644 utils.py create mode 100644 小实验/t.json create mode 100644 小实验/网页访问gpt-4.py create mode 100644 小实验/网页访问gpt-4——上传文件.py create mode 100644 论文信息爬取（题目、期刊、日期、摘要、关键词）_1.py diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..98c62e4 --- /dev/null +++ b/.gitignore @@ -0,0 +1,4 @@ +__pycache__ +.idea +paper_download/ +te_u/paper_down_load/ECCV_2022/ \ No newline at end of file diff --git a/README.md b/README.md new file mode 100644 index 0000000..e69de29 diff --git a/main.py b/main.py new file mode 100644 index 0000000..5e350da --- /dev/null +++ b/main.py @@ -0,0 +1,176 @@ +import gradio as gr +import os + +from te_u.arxiv import get_news_from_arxiv +# +# os.environ['http_proxy'] = '127.0.0.1:7890' +# os.environ['https_proxy'] = '127.0.0.1:7890' + +from utils import get_news, get_clouds +from gradio_pdf import PDF + +current_pdf_file = None +news = [] +choose_news = [] + +with gr.Blocks() as demo: + with gr.Row(): + with gr.Column(scale=20): + gr.HTML("""

科研情报

""") + with gr.Column(scale=1, min_width=100): + gr.HTML( + """
""" + ) + gr.HTML( + """
Created by 朱瑞
""" + ) + + with gr.Tabs(elem_classes="tab-buttons") as tabs: + with gr.TabItem("科研文献分析"): + with gr.Row(): + with gr.Accordion("文献采集区", open=True, ) as area_news_get_fn: + keywords = gr.Dropdown(choices=["对抗攻击", "knowledge graph", "认知智能与先进计算", "电磁空间感知与利用", "信息安全与攻防博弈"], + value="对抗攻击", label="关键词", show_label=True) + source = gr.Dropdown(choices=["知网", "arxiv"], value="知网", label="数据源", show_label=True) + num = gr.Slider(1, 100, value=10, label="采集条数", step=1) + news_get = gr.Button("获取论文", variant='primary') + + with gr.Row(): + with gr.Accordion("文献标记分析区", open=True, elem_id="news-panel") as news_get_fn: + chosen_news = gr.CheckboxGroup(choices=[item['name'] for item in news], label="需要进行操作的文献") + + with gr.Row(): + news_mark = gr.Button("标记文献") + news_all_mark = gr.Button("全部标记", variant='primary') + + + def recover_news_by_choose(news_titles): + select_news = [] + global news + + for news_title in news_titles: + for i in news: + if news_title == i['name']: + new_i = i + select_news.append(new_i) + break + + return select_news + + + def mark_new(titles): + global choose_news + mark_news = recover_news_by_choose(titles) + choose_news = mark_news + + + def get_news_temp(num, keywords, source): + """ 获取临时的文献 """ + global news + results = [] + if source == "知网": + results = get_news(num, keywords) + elif source == "arxiv": + results = get_news_from_arxiv(num, keywords) + + news.extend(results) + return gr.CheckboxGroup(choices=[item['name'] for item in news], label="需要进行操作的文献") + + + def mark_all_new(): + global news + global choose_news + choose_news = news + return gr.CheckboxGroup(choices=[item['name'] for item in news], value=[item['name'] for item in news], label="需要进行操作的文献") + + + news_get.click(get_news_temp, inputs=[num, keywords, source], outputs=[chosen_news]) + news_mark.click(mark_new, inputs=[chosen_news]) + news_all_mark.click(mark_all_new, outputs=[chosen_news]) + + with gr.TabItem("科研文献获取"): + with gr.Row(): + with gr.Accordion("功能区", open=True, ) as area_news_analyse_fn: + with gr.Row(): + ci_yun_by_title = gr.Button("题目词云", variant='primary') + ci_yun_by_abstract = gr.Button("摘要词云", variant='primary') + with gr.Row(): + with gr.Accordion("结果展示区", open=True, ) as area_news_result_fn: + result_place = gr.Image() + + + def g_ci_yun_by_title(): + global choose_news + word_list = [c["name"] for c in choose_news] + pic = get_clouds(word_list) + return pic + + + def g_ci_yun_by_abstract(): + global choose_news + word_list = [c["abstract"] for c in choose_news] + pic = get_clouds(word_list) + return pic + + + ci_yun_by_title.click(g_ci_yun_by_title, outputs=[result_place]) + ci_yun_by_abstract.click(g_ci_yun_by_abstract, outputs=[result_place]) + + with gr.TabItem("会议论文查看"): + with gr.Row(): + with gr.Column(scale=1): + with gr.Row(): + # gr.Label("会议名称") + conf_name = gr.Dropdown(choices=["ECCV2022", "ECCV2020", "CVPR2024"], value="ECCV2022", label="会议名称", show_label=True) + conf_button = gr.Button("查看会议论文", variant='primary') + dataframe = gr.Dataframe(headers=["论文名称"], col_count=(1, "fixed"), type='array', height=800) + with gr.Row(): + look_input = gr.Textbox(placeholder="关键词检索", label="关键词过滤") + filter_button = gr.Button("过滤") + # up_button = gr.Button("加载") + + with gr.Column(scale=2): + pdf = PDF(label="Upload a PDF", interactive=True, height=1000) + + + # name = gr.Textbox(show_label=False) + # pdf.upload(lambda f: f, pdf, name) + + def up_load(): + global current_pdf_file + n = r"D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_2022\main_paper\3d-siamese-transformer-network-for-single-object-tracking-on-point-clouds_ECCV_2022.pdf" + current_pdf_file = n + return n + + + def load_conf_list(conf_name): + if conf_name == "ECCV2022": + root_dir = r"D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_2022\main_paper" + return [[i] for i in os.listdir(root_dir)] + + + def look_dataframe(evt: gr.SelectData): + global current_pdf_file + if evt.value: + root_dir = r"D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_2022\main_paper" + n = os.path.join(root_dir, evt.value) + if os.path.exists(n): + current_pdf_file = n + return current_pdf_file + + + def filter_by_word(words, paper_list): + word_list = words.strip().split() + paper_list_filter = [p[0] for p in paper_list] + for word in word_list: + paper_list_filter = [p for p in paper_list_filter if word in p] + return [[p] for p in paper_list_filter] + + + filter_button.click(filter_by_word, inputs=[look_input, dataframe], outputs=[dataframe]) + dataframe.select(look_dataframe, inputs=None, outputs=[pdf]) + conf_button.click(load_conf_list, inputs=[conf_name], outputs=[dataframe]) + # up_button.click(up_load, inputs=None, outputs=[pdf])s + +if __name__ == '__main__': + demo.queue().launch(inbrowser=True, server_name='127.0.0.1', server_port=23223) diff --git a/method3_dict.txt b/method3_dict.txt new file mode 100644 index 0000000..d550e6a --- /dev/null +++ b/method3_dict.txt @@ -0,0 +1,20 @@ +_trad_ 12.215113775321004 +task 11.808329224486352 +step 10.115128616689704 +thought 9.468294108747731 +performance 7.91112148935495 +agent 7.908585185590241 +demonstration 7.695334786087041 +retrieval 7.60209065815528 +method 7.186012901911181 +trajectory 6.258998528039508 +information 5.995282554194667 +_synapse_ 5.572552304074627 +relevant 5.527015778258248 +example 5.080665441372099 +reason 4.676097441406382 +_react_ 4.570513969848461 +baseline 4.479754027332443 +prompt 4.395961022082388 +achieve 4.296215825920176 +current 4.284028839203101 diff --git a/result.json b/result.json new file mode 100644 index 0000000..91133c2 --- /dev/null +++ b/result.json @@ -0,0 +1 @@ +[{"name": "\u9488\u5bf9\u7535\u529bCPS\u6570\u636e\u9a71\u52a8\u7b97\u6cd5\u5bf9\u6297\u653b\u51fb\u7684\u9632\u5fa1\u65b9\u6cd5", "authors": ["\u6731\u536b\u5e731", "\u6c64\u59552", "\u9b4f\u5174\u614e3", "\u5218\u589e\u7a372"], "affiliations": ["1. \u56fd\u7f51\u6c5f\u82cf\u7701\u7535\u529b\u6709\u9650\u516c\u53f8", "2. \u4e1c\u5357\u5927\u5b66\u7535\u6c14\u5de5\u7a0b\u5b66\u9662", "3. \u5357\u745e\u96c6\u56e2\u6709\u9650\u516c\u53f8(\u56fd\u7f51\u7535\u529b\u79d1\u5b66\u7814\u7a76\u9662\u6709\u9650\u516c\u53f8)"], "abstract": "\u5927\u89c4\u6a21\u7535\u529b\u7535\u5b50\u8bbe\u5907\u7684\u63a5\u5165\u4e3a\u7cfb\u7edf\u5f15\u5165\u4e86\u6570\u91cf\u5e9e\u5927\u7684\u5f3a\u975e\u7ebf\u6027\u91cf\u6d4b/\u63a7\u5236\u8282\u70b9\uff0c\u4f7f\u5f97\u4f20\u7edf\u7535\u529b\u7cfb\u7edf\u9010\u6e10\u8f6c\u53d8\u4e3a\u7535\u529b\u4fe1\u606f\u7269\u7406\u7cfb\u7edf\uff08cyber-physical system\uff0c CPS\uff09\uff0c\u8bb8\u591a\u539f\u672c\u5e94\u7528\u6a21\u578b\u9a71\u52a8\u65b9\u6cd5\u89e3\u51b3\u7684\u7cfb\u7edf\u95ee\u9898\u4e0d\u5f97\u4e0d\u56e0\u7ef4\u5ea6\u707e\u96be\u7b49\u5c40\u9650\u8f6c\u800c\u91c7\u53d6\u6570\u636e\u9a71\u52a8\u7b97\u6cd5\u8fdb\u884c\u5206\u6790\u3002\u7136\u800c\uff0c\u6570\u636e\u9a71\u52a8\u7b97\u6cd5\u81ea\u8eab\u7684\u7f3a\u9677\u4e3a\u7cfb\u7edf\u7684\u5b89\u5168\u7a33\u5b9a\u8fd0\u884c\u5f15\u5165\u4e86\u65b0\u7684\u98ce\u9669\uff0c\u653b\u51fb\u8005\u53ef\u4ee5\u5bf9\u5176\u52a0\u4ee5\u5229\u7528\uff0c\u53d1\u8d77\u53ef\u80fd\u5f15\u53d1\u7cfb\u7edf\u505c\u7535\u751a\u81f3\u5931\u7a33\u7684\u5bf9\u6297\u653b\u51fb\u3002\u9488\u5bf9\u7535\u529bCPS\u4e2d\u6570\u636e\u9a71\u52a8\u7b97\u6cd5\u53ef\u80fd\u906d\u53d7\u7684\u5bf9\u6297\u653b\u51fb\uff0c\u4ece\u5f02\u5e38\u6570\u636e\u5254\u9664\u4e0e\u6062\u590d\u3001\u7b97\u6cd5\u6f0f\u6d1e\u6316\u6398\u4e0e\u4f18\u5316\u3001\u7b97\u6cd5\u81ea\u8eab\u53ef\u89e3\u91ca\u6027\u63d0\u53473\u4e2a\u65b9\u9762\uff0c\u63d0\u51fa\u4e86\u5bf9\u5e94\u7684\u9632\u5fa1\u65b9\u6cd5\uff1a\u5f02\u5e38\u6570\u636e\u8fc7\u6ee4\u5668\u3001\u57fa\u4e8eGAN\u7684\u6f0f\u6d1e\u6316\u6398\u4e0e\u4f18\u5316\u65b9\u6cd5\u3001\u6570\u636e-\u77e5\u8bc6\u878d\u5408\u6a21\u578b\u53ca\u5176\u8bad\u7ec3\u65b9\u6cd5\uff0c\u5e76\u7ecf\u7b97\u4f8b\u5206\u6790\u9a8c\u8bc1\u4e86\u6240\u63d0\u65b9\u6cd5\u7684\u6709\u6548\u6027\u3002"}, {"name": "\u878d\u5408\u98ce\u683c\u8fc1\u79fb\u7684\u5bf9\u6297\u6837\u672c\u751f\u6210\u65b9\u6cd5", "authors": ["\u4e8e\u632f\u534e", "\u6bb7\u6b63", "\u53f6\u9e25", "\u4e1b\u65ed\u4e9a"], "affiliations": ["\u897f\u5b89\u79d1\u6280\u5927\u5b66\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u5b66\u9662"], "abstract": "\u9488\u5bf9\u73b0\u6709\u9762\u5411\u76ee\u6807\u68c0\u6d4b\u7684\u5bf9\u6297\u6837\u672c\u751f\u6210\u65b9\u6cd5\u6cdb\u5316\u80fd\u529b\u5f31\u7684\u95ee\u9898\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u878d\u5408\u98ce\u683c\u8fc1\u79fb\u7684\u5bf9\u6297\u6837\u672c\u751f\u6210\u65b9\u6cd5\u3002\u9996\u5148\u63d0\u51fa\u4e00\u79cd\u65b0\u7684\u5bf9\u6297\u8865\u4e01\u751f\u6210\u65b9\u6cd5\uff0c\u4f7f\u7528\u98ce\u683c\u8fc1\u79fb\u65b9\u6cd5\u5c06\u98ce\u683c\u56fe\u50cf\u4e0d\u540c\u5c42\u6b21\u7279\u5f81\u63d0\u53d6\u5e76\u878d\u5408\uff0c\u751f\u6210\u65e0\u660e\u663e\u7269\u4f53\u7279\u5f81\u4e14\u7eb9\u7406\u4e30\u5bcc\u7684\u5bf9\u6297\u8865\u4e01\uff1b\u7136\u540e\u5229\u7528\u68af\u5ea6\u7c7b\u6fc0\u6d3b\u6620\u5c04\u65b9\u6cd5\u751f\u6210\u76ee\u6807\u7684\u7279\u5f81\u70ed\u56fe\uff0c\u5c06\u76ee\u6807\u4e0d\u540c\u533a\u57df\u5728\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u4e2d\u7684\u5173\u952e\u7a0b\u5ea6\u8fdb\u884c\u53ef\u89c6\u5316\u8868\u793a\uff1b\u6700\u540e\u6784\u5efa\u4e00\u79cd\u70ed\u56fe\u5f15\u5bfc\u673a\u5236\uff0c\u5f15\u5bfc\u5bf9\u6297\u8865\u4e01\u5728\u653b\u51fb\u76ee\u6807\u7684\u5173\u952e\u4f4d\u7f6e\u8fdb\u884c\u653b\u51fb\u4ee5\u63d0\u9ad8\u5176\u6cdb\u5316\u80fd\u529b\uff0c\u751f\u6210\u6700\u7ec8\u5bf9\u6297\u6837\u672c\u3002\u4e3a\u4e86\u9a8c\u8bc1\u6240\u63d0\u65b9\u6cd5\u7684\u6027\u80fd\uff0c\u5728DroNet\u5ba4\u5916\u6570\u636e\u96c6\u4e0a\u8fdb\u884c\u4e86\u5b9e\u9a8c\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u6240\u63d0\u65b9\u6cd5\u9488\u5bf9\u5355\u9636\u6bb5\u76ee\u6807\u68c0\u6d4b\u6a21\u578bYOLOv5\u751f\u6210\u7684\u5bf9\u6297\u6837\u672c\uff0c\u653b\u51fb\u6210\u529f\u7387\u53ef\u8fbe84.07%\uff0c\u5c06\u5176\u5e94\u7528\u4e8e\u653b\u51fb\u4e24\u9636\u6bb5\u76ee\u6807\u68c0\u6d4b\u6a21\u578bFaster R-CNN\u65f6\uff0c\u653b\u51fb\u6210\u529f\u7387\u4ecd\u4fdd\u6301\u572867.65%\u3002\u4e0e\u6240\u5bf9\u6bd4\u7684\u4e3b\u6d41\u65b9\u6cd5\u76f8\u6bd4\uff0c\u6240\u63d0\u65b9\u6cd5\u751f\u6210\u7684\u5bf9\u6297\u6837\u672c\u653b\u51fb\u6548\u679c\u8f83\u597d\uff0c\u800c\u4e14\u5177\u6709\u826f\u597d\u7684\u6cdb\u5316\u80fd\u529b\u3002"}, {"name": "\u57fa\u4e8eSE-AdvGAN\u7684\u56fe\u50cf\u5bf9\u6297\u6837\u672c\u751f\u6210\u65b9\u6cd5\u7814\u7a76", "authors": ["\u8d75\u5b8f", "\u5b8b\u99a5\u8363", "\u674e\u6587\u6539"], "affiliations": ["\u5170\u5dde\u7406\u5de5\u5927\u5b66\u8ba1\u7b97\u673a\u4e0e\u901a\u4fe1\u5b66\u9662"], "abstract": "\u5bf9\u6297\u6837\u672c\u662f\u8bc4\u4f30\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u9c81\u68d2\u6027\u548c\u63ed\u793a\u5176\u6f5c\u5728\u5b89\u5168\u9690\u60a3\u7684\u91cd\u8981\u624b\u6bb5\u3002\u57fa\u4e8e\u751f\u6210\u5bf9\u6297\u7f51\u7edc(GAN)\u7684\u5bf9\u6297\u6837\u672c\u751f\u6210\u65b9\u6cd5(AdvGAN)\u5728\u751f\u6210\u56fe\u50cf\u5bf9\u6297\u6837\u672c\u65b9\u9762\u53d6\u5f97\u663e\u8457\u8fdb\u5c55\uff0c\u4f46\u8be5\u65b9\u6cd5\u751f\u6210\u7684\u6270\u52a8\u7a00\u758f\u6027\u4e0d\u8db3\u4e14\u5e45\u5ea6\u8f83\u5927\uff0c\u5bfc\u81f4\u5bf9\u6297\u6837\u672c\u7684\u771f\u5b9e\u6027\u8f83\u4f4e\u3002\u4e3a\u89e3\u51b3\u8fd9\u4e00\u95ee\u9898\uff0c\u57fa\u4e8eAdvGAN\u63d0\u51fa\u4e00\u79cd\u6539\u8fdb\u7684\u56fe\u50cf\u5bf9\u6297\u6837\u672c\u751f\u6210\u65b9\u6cd5Squeeze-and-Excitation-AdvGAN(SE-AdvGAN)\u3002SE-AdvGAN\u901a\u8fc7\u6784\u9020SE\u6ce8\u610f\u529b\u751f\u6210\u5668\u548cSE\u6b8b\u5dee\u5224\u522b\u5668\u63d0\u9ad8\u6270\u52a8\u7684\u7a00\u758f\u6027\u3002SE\u6ce8\u610f\u529b\u751f\u6210\u5668\u7528\u4e8e\u63d0\u53d6\u56fe\u50cf\u5173\u952e\u7279\u5f81\u9650\u5236\u6270\u52a8\u751f\u6210\u4f4d\u7f6e\uff0cSE\u6b8b\u5dee\u5224\u522b\u5668\u6307\u5bfc\u751f\u6210\u5668\u907f\u514d\u751f\u6210\u65e0\u5173\u6270\u52a8\u3002\u540c\u65f6\uff0c\u5728SE\u6ce8\u610f\u529b\u751f\u6210\u5668\u7684\u635f\u5931\u51fd\u6570\u4e2d\u52a0\u5165\u4ee5\u25a0\u8303\u6570\u4e3a\u57fa\u51c6\u7684\u8fb9\u754c\u635f\u5931\u4ee5\u9650\u5236\u6270\u52a8\u7684\u5e45\u5ea6\uff0c\u4ece\u800c\u63d0\u9ad8\u5bf9\u6297\u6837\u672c\u7684\u771f\u5b9e\u6027\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u5728\u767d\u76d2\u653b\u51fb\u573a\u666f\u4e0b\uff0cSE-AdvGAN\u76f8\u8f83\u4e8e\u73b0\u6709\u65b9\u6cd5\u751f\u6210\u7684\u5bf9\u6297\u6837\u672c\u6270\u52a8\u7a00\u758f\u6027\u9ad8\u3001\u5e45\u5ea6\u5c0f\uff0c\u5e76\u4e14\u5728\u4e0d\u540c\u76ee\u6807\u6a21\u578b\u4e0a\u5747\u53d6\u5f97\u66f4\u597d\u7684\u653b\u51fb\u6548\u679c\uff0c\u8bf4\u660eSE-AdvGAN\u751f\u6210\u7684\u9ad8\u8d28\u91cf\u5bf9\u6297\u6837\u672c\u53ef\u4ee5\u66f4\u6709\u6548\u5730\u8bc4\u4f30\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u7684\u9c81\u68d2\u6027\u3002"}, {"name": "\u9762\u5411\u6f0f\u6d1e\u68c0\u6d4b\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60\u5f0f\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5", "authors": ["\u9648\u601d\u71361,2", "\u5434\u656c\u5f811,3", "\u51cc\u79651", "\u7f57\u5929\u60a61", "\u5218\u9553\u715c1,2", "\u6b66\u5ef6\u519b1,3"], "affiliations": ["1. \u4e2d\u56fd\u79d1\u5b66\u9662\u8f6f\u4ef6\u7814\u7a76\u6240\u667a\u80fd\u8f6f\u4ef6\u7814\u7a76\u4e2d\u5fc3", "2. \u4e2d\u56fd\u79d1\u5b66\u9662\u5927\u5b66", "3. \u8ba1\u7b97\u673a\u79d1\u5b66\u56fd\u5bb6\u91cd\u70b9\u5b9e\u9a8c\u5ba4(\u4e2d\u56fd\u79d1\u5b66\u9662\u8f6f\u4ef6\u7814\u7a76\u6240)"], "abstract": "\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u4ee3\u7801\u6f0f\u6d1e\u68c0\u6d4b\u6a21\u578b\u56e0\u5176\u68c0\u6d4b\u6548\u7387\u9ad8\u548c\u7cbe\u5ea6\u51c6\u7684\u4f18\u52bf,\u9010\u6b65\u6210\u4e3a\u68c0\u6d4b\u8f6f\u4ef6\u6f0f\u6d1e\u7684\u91cd\u8981\u65b9\u6cd5,\u5e76\u5728\u4ee3\u7801\u6258\u7ba1\u5e73\u53f0Github\u7684\u4ee3\u7801\u5ba1\u8ba1\u670d\u52a1\u4e2d\u53d1\u6325\u91cd\u8981\u4f5c\u7528.\u7136\u800c,\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u5df2\u88ab\u8bc1\u660e\u5bb9\u6613\u53d7\u5230\u5bf9\u6297\u653b\u51fb\u7684\u5e72\u6270,\u8fd9\u5bfc\u81f4\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u6f0f\u6d1e\u68c0\u6d4b\u6a21\u578b\u5b58\u5728\u906d\u53d7\u653b\u51fb\u3001\u964d\u4f4e\u68c0\u6d4b\u51c6\u786e\u7387\u7684\u98ce\u9669.\u56e0\u6b64,\u6784\u5efa\u9488\u5bf9\u6f0f\u6d1e\u68c0\u6d4b\u6a21\u578b\u7684\u5bf9\u6297\u653b\u51fb\u4e0d\u4ec5\u53ef\u4ee5\u53d1\u6398\u6b64\u7c7b\u6a21\u578b\u7684\u5b89\u5168\u7f3a\u9677,\u800c\u4e14\u6709\u52a9\u4e8e\u8bc4\u4f30\u6a21\u578b\u7684\u9c81\u68d2\u6027,\u8fdb\u800c\u901a\u8fc7\u76f8\u5e94\u7684\u65b9\u6cd5\u63d0\u5347\u6a21\u578b\u6027\u80fd.\u4f46\u73b0\u6709\u7684\u9762\u5411\u6f0f\u6d1e\u68c0\u6d4b\u6a21\u578b\u7684\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\u4f9d\u8d56\u4e8e\u901a\u7528\u7684\u4ee3\u7801\u8f6c\u6362\u5de5\u5177,\u5e76\u672a\u63d0\u51fa\u9488\u5bf9\u6027\u7684\u4ee3\u7801\u6270\u52a8\u64cd\u4f5c\u548c\u51b3\u7b56\u7b97\u6cd5,\u56e0\u6b64\u96be\u4ee5\u751f\u6210\u6709\u6548\u7684\u5bf9\u6297\u6837\u672c,\u4e14\u5bf9\u6297\u6837\u672c\u7684\u5408\u6cd5\u6027\u4f9d\u8d56\u4e8e\u4eba\u5de5\u68c0\u67e5.\u9488\u5bf9\u4e0a\u8ff0\u95ee\u9898,\u63d0\u51fa\u4e86\u4e00\u79cd\u9762\u5411\u6f0f\u6d1e\u68c0\u6d4b\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60\u5f0f\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5.\u8be5\u65b9\u6cd5\u9996\u5148\u8bbe\u8ba1\u4e86\u4e00\u7cfb\u5217\u8bed\u4e49\u7ea6\u675f\u4e14\u6f0f\u6d1e\u4fdd\u7559\u7684\u4ee3\u7801\u6270\u52a8\u64cd\u4f5c\u4f5c\u4e3a\u6270\u52a8\u96c6\u5408;\u5176\u6b21,\u5c06\u5177\u5907\u6f0f\u6d1e\u7684\u4ee3\u7801\u6837\u672c\u4f5c\u4e3a\u8f93\u5165,\u5229\u7528\u5f3a\u5316\u5b66\u4e60\u6a21\u578b\u9009\u53d6\u5177\u4f53\u7684\u6270\u52a8\u64cd\u4f5c\u5e8f\u5217;\u6700\u540e,\u6839\u636e\u4ee3\u7801\u6837\u672c\u7684\u8bed\u6cd5\u6811\u8282\u70b9\u7c7b\u578b\u5bfb\u627e\u6270\u52a8\u7684\u6f5c\u5728\u4f4d\u7f6e,\u8fdb\u884c\u4ee3\u7801\u8f6c\u6362,\u4ece\u800c\u751f\u6210\u5bf9\u6297\u6837\u672c.\u57fa\u4e8eSARD\u548cNVD\u6784\u5efa\u4e86\u4e24\u4e2a\u5b9e\u9a8c\u6570\u636e\u96c6,\u517114278\u4e2a\u4ee3\u7801\u6837\u672c,\u5e76\u4ee5\u6b64\u8bad\u7ec3\u4e864\u4e2a\u5177\u5907\u4e0d\u540c\u7279\u70b9\u7684\u6f0f\u6d1e\u68c0\u6d4b\u6a21\u578b\u4f5c\u4e3a\u653b\u51fb\u76ee\u6807.\u9488\u5bf9\u6bcf\u4e2a\u76ee\u6807\u6a21\u578b,\u8bad\u7ec3\u4e86\u4e00\u4e2a\u5f3a\u5316\u5b66\u4e60\u7f51\u7edc\u8fdb\u884c\u5bf9\u6297\u653b\u51fb.\u7ed3\u679c\u663e\u793a,\u8be5\u653b\u51fb\u65b9\u6cd5\u5bfc\u81f4\u6a21\u578b\u7684\u53ec\u56de\u7387\u964d\u4f4e\u4e8674.34%,\u653b\u51fb\u6210\u529f\u7387\u8fbe\u523096.71%,\u76f8\u8f83\u57fa\u7ebf\u65b9\u6cd5,\u653b\u51fb\u6210\u529f\u7387\u5e73\u5747\u63d0\u5347\u4e8668.76%.\u5b9e\u9a8c\u8bc1\u660e\u4e86\u5f53\u524d\u7684\u6f0f\u6d1e\u68c0\u6d4b\u6a21\u578b\u5b58\u5728\u88ab\u653b\u51fb\u7684\u98ce\u9669,\u9700\u8981\u8fdb\u4e00\u6b65\u7814\u7a76\u63d0\u5347\u6a21\u578b\u7684\u9c81\u68d2\u6027. "}] \ No newline at end of file diff --git a/result1.json b/result1.json new file mode 100644 index 0000000..1d73ce5 --- /dev/null +++ b/result1.json @@ -0,0 +1,518 @@ +[ + { + "name": "\u57fa\u4e8e\u6570\u636e\u589e\u5f3a\u548c\u6807\u7b7e\u566a\u58f0\u7684\u5feb\u901f\u5bf9\u6297\u8bad\u7ec3\u65b9\u6cd5", + "authors": [ + "\u5b8b\u9038\u98de", + "\u67f3\u6bc5" + ], + "affiliations": [ + "\u5e7f\u4e1c\u5de5\u4e1a\u5927\u5b66\u8ba1\u7b97\u673a\u5b66\u9662" + ], + "abstract": "\u5bf9\u6297\u8bad\u7ec3\u662f\u4fdd\u62a4\u5206\u7c7b\u6a21\u578b\u514d\u53d7\u5bf9\u6297\u6027\u653b\u51fb\u7684\u6709\u6548\u9632\u5fa1\u65b9\u6cd5\u3002\u7136\u800c\uff0c\u7531\u4e8e\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u751f\u6210\u5f3a\u5bf9\u6297\u6837\u672c\u7684\u9ad8\u6210\u672c\uff0c\u53ef\u80fd\u9700\u8981\u6570\u91cf\u7ea7\u7684\u989d\u5916\u8bad\u7ec3\u65f6\u95f4\u3002\u4e3a\u4e86\u514b\u670d\u8fd9\u4e00\u9650\u5236\uff0c\u57fa\u4e8e\u5355\u6b65\u653b\u51fb\u7684\u5feb\u901f\u5bf9\u6297\u8bad\u7ec3\u5df2\u88ab\u63a2\u7d22\u3002\u4ee5\u5f80\u7684\u5de5\u4f5c\u4ece\u6837\u672c\u521d\u59cb\u5316\u3001\u635f\u5931\u6b63\u5219\u5316\u548c\u8bad\u7ec3\u7b56\u7565\u7b49\u4e0d\u540c\u89d2\u5ea6\u5bf9\u5feb\u901f\u5bf9\u6297\u8bad\u7ec3\u8fdb\u884c\u4e86\u6539\u8fdb\u3002\u7136\u800c\uff0c\u5728\u5904\u7406\u5927\u6270\u52a8\u9884\u7b97\u65f6\u9047\u5230\u4e86\u707e\u96be\u6027\u8fc7\u62df\u5408\u3002\u57fa\u4e8e\u6570\u636e\u589e\u5f3a\u4e0e\u6807\u7b7e\u566a\u58f0\u7684\u5feb\u901f\u5bf9\u6297\u8bad\u7ec3\u65b9\u6cd5\u88ab\u63d0\u51fa\uff0c\u4ee5\u89e3\u51b3\u6b64\u56f0\u96be\u3002\u521d\u59cb\u9636\u6bb5\uff0c\u5bf9\u539f\u59cb\u6837\u672c\u6267\u884c\u591a\u79cd\u56fe\u50cf\u8f6c\u6362\uff0c\u5e76\u5f15\u5165\u968f\u673a\u566a\u58f0\u4ee5\u5b9e\u65bd\u6570\u636e\u589e\u5f3a\uff1b\u63a5\u7740\uff0c\u5c11\u91cf\u6807\u7b7e\u566a\u58f0\u88ab\u6ce8\u5165\uff1b\u7136\u540e\u4f7f\u7528\u589e\u5f3a\u7684\u6570\u636e\u751f\u6210\u5bf9\u6297\u6837\u672c\u7528\u4e8e\u6a21\u578b\u8bad\u7ec3\uff1b\u6700\u540e\uff0c\u6839\u636e\u5bf9\u6297\u9c81\u68d2\u6027\u6d4b\u8bd5\u7ed3\u679c\u81ea\u9002\u5e94\u5730\u8c03\u6574\u6807\u7b7e\u566a\u58f0\u7387\u3002\u5728CIFAR-10\u3001CIFAR-100\u6570\u636e\u96c6\u4e0a\u7684\u5168\u9762\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u76f8\u8f83\u4e8eFGSM-MEP\uff0c\u6240\u63d0\u65b9\u6cd5\u5728\u5927\u6270\u52a8\u9884\u7b97\u6761\u4ef6\u4e0b\uff0c\u5728\u4e24\u4e2a\u6570\u636e\u96c6\u4e0a\u7684AA\u4e0a\u5206\u522b\u63d0\u5347\u4e864.63\u548c5.38\u4e2a\u767e\u5206\u70b9\u3002\u7ecf\u5b9e\u9a8c\u8bc1\u660e\uff0c\u65b0\u63d0\u51fa\u7684\u65b9\u6848\u53ef\u4ee5\u6709\u6548\u5730\u5904\u7406\u5927\u7684\u6270\u52a8\u9884\u7b97\u4e0b\u707e\u96be\u6027\u8fc7\u62df\u5408\u95ee\u9898\uff0c\u5e76\u663e\u8457\u589e\u5f3a\u6a21\u578b\u7684\u5bf9\u6297\u9c81\u68d2\u6027\u3002" + }, + { + "name": "\u57fa\u4e8e\u6761\u4ef6\u6269\u6563\u6a21\u578b\u7684\u56fe\u50cf\u5206\u7c7b\u5bf9\u6297\u6837\u672c\u9632\u5fa1\u65b9\u6cd5", + "authors": [ + "\u9648\u5b50\u6c11", + "\u5173\u5fd7\u6d9b" + ], + "affiliations": [ + "\u534e\u5317\u7535\u529b\u5927\u5b66\u63a7\u5236\u4e0e\u8ba1\u7b97\u673a\u5b66\u9662" + ], + "abstract": "\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5728\u56fe\u50cf\u5206\u7c7b\u7b49\u9886\u57df\u53d6\u5f97\u4ee4\u4eba\u5370\u8c61\u6df1\u523b\u7684\u7ed3\u679c\uff0c\u4f46\u662f\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5bb9\u6613\u53d7\u5230\u5bf9\u6297\u6837\u672c\u7684\u5e72\u6270\u5a01\u80c1\uff0c\u653b\u51fb\u8005\u901a\u8fc7\u5bf9\u6297\u6837\u672c\u5236\u4f5c\u7b97\u6cd5\uff0c\u7cbe\u5fc3\u8bbe\u8ba1\u5fae\u5c0f\u6270\u52a8\uff0c\u6784\u9020\u8089\u773c\u96be\u4ee5\u5206\u8fa8\u5374\u80fd\u5f15\u53d1\u6a21\u578b\u8bef\u5206\u7c7b\u7684\u5bf9\u6297\u6837\u672c\uff0c\u7ed9\u56fe\u50cf\u5206\u7c7b\u7b49\u6df1\u5ea6\u5b66\u4e60\u5e94\u7528\u5e26\u6765\u4e25\u91cd\u7684\u5b89\u5168\u9690\u60a3\u3002\u4e3a\u63d0\u5347\u56fe\u50cf\u5206\u7c7b\u6a21\u578b\u7684\u9c81\u68d2\u6027\uff0c\u672c\u6587\u5229\u7528\u6761\u4ef6\u6269\u6563\u6a21\u578b\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u7efc\u5408\u5bf9\u6297\u6837\u672c\u68c0\u6d4b\u548c\u5bf9\u6297\u6837\u672c\u51c0\u5316\u7684\u5bf9\u6297\u6837\u672c\u9632\u5fa1\u65b9\u6cd5\u3002\u5728\u4e0d\u4fee\u6539\u76ee\u6807\u6a21\u578b\u7684\u57fa\u7840\u4e0a\uff0c\u68c0\u6d4b\u5e76\u51c0\u5316\u5bf9\u6297\u6837\u672c\uff0c\u63d0\u5347\u76ee\u6807\u6a21\u578b\u9c81\u68d2\u6027\u3002\u672c\u65b9\u6cd5\u5305\u62ec\u5bf9\u6297\u6837\u672c\u68c0\u6d4b\u548c\u5bf9\u6297\u6837\u672c\u51c0\u5316\u4e24\u4e2a\u6a21\u5757\u3002\u5bf9\u4e8e\u5bf9\u6297\u6837\u672c\u68c0\u6d4b\uff0c\u91c7\u7528\u4e0d\u4e00\u81f4\u6027\u589e\u5f3a\uff0c\u901a\u8fc7\u8bad\u7ec3\u4e00\u4e2a\u878d\u5165\u76ee\u6807\u6a21\u578b\u9ad8\u7ef4\u7279\u5f81\u548c\u56fe\u7247\u57fa\u672c\u7279\u5f81\u7684\u56fe\u50cf\u4fee\u590d\u6a21\u578b\uff0c\u6bd4\u8f83\u521d\u59cb\u8f93\u5165\u548c\u4fee\u590d\u7ed3\u679c\u7684\u4e0d\u4e00\u81f4\u6027\uff0c\u68c0\u6d4b\u5bf9\u6297\u6837\u672c\uff1b\u5bf9\u4e8e\u5bf9\u6297\u6837\u672c\u51c0\u5316\uff0c\u91c7\u7528\u7aef\u5230\u7aef\u7684\u5bf9\u6297\u6837\u672c\u51c0\u5316\u65b9\u5f0f\uff0c\u5728\u53bb\u566a\u6a21\u578b\u6267\u884c\u8fc7\u7a0b\u4e2d\u52a0\u5165\u56fe\u7247\u4f2a\u5f71\uff0c\u5b9e\u73b0\u5bf9\u6297\u6837\u672c\u51c0\u5316\u3002\u5728\u4fdd\u8bc1\u76ee\u6807\u6a21\u578b\u7cbe\u5ea6\u7684\u524d\u63d0\u4e0b\uff0c\u5728\u76ee\u6807\u6a21\u578b\u524d\u589e\u52a0\u5bf9\u6297\u6837\u672c\u68c0\u6d4b\u548c\u51c0\u5316\u6a21\u5757\uff0c\u6839\u636e\u68c0\u6d4b\u7ed3\u679c\uff0c\u9009\u53d6\u76f8\u5e94\u7684\u51c0\u5316\u7b56\u7565\uff0c\u4ece\u800c\u6d88\u9664\u5bf9\u6297\u6837\u672c\uff0c\u63d0\u5347\u76ee\u6807\u6a21\u578b\u7684\u9c81\u68d2\u6027\u3002\u5728CIFAR10\u6570\u636e\u96c6\u548cCIFAR100\u6570\u636e\u96c6\u4e0a\u4e0e\u73b0\u6709\u65b9\u6cd5\u8fdb\u884c\u5bf9\u6bd4\u5b9e\u9a8c\u3002\u5bf9\u4e8e\u6270\u52a8\u8f83\u5c0f\u7684\u5bf9\u6297\u6837\u672c\uff0c\u672c\u65b9\u6cd5\u7684\u68c0\u6d4b\u7cbe\u5ea6\u6bd4Argos\u65b9\u6cd5\u63d0\u5347\u4e865-9\u4e2a\u767e\u5206\u70b9\uff1b\u76f8\u6bd4\u4e8eADP\u65b9\u6cd5\uff0c\u672c\u65b9\u6cd5\u5728\u9762\u5bf9\u4e0d\u540c\u79cd\u7c7b\u5bf9\u6297\u6837\u672c\u65f6\u9632\u5fa1\u6548\u679c\u66f4\u7a33\u5b9a\uff0c\u4e14\u5728BPDA\u653b\u51fb\u4e0b\uff0c\u672c\u65b9\u6cd5\u7684\u5bf9\u6297\u6837\u672c\u51c0\u5316\u6548\u679c\u8f83ADP\u63d0\u5347\u4e861.3%\u3002 " + }, + { + "name": "\u57fa\u4e8e\u63a9\u6a21\u63d0\u53d6\u7684SAR\u56fe\u50cf\u5bf9\u6297\u6837\u672c\u751f\u6210\u65b9\u6cd5", + "authors": [ + "\u7ae0\u575a\u6b661", + "\u80fd\u8c6a1", + "\u674e\u67701", + "\u94b1\u5efa\u534e2" + ], + "affiliations": [ + "1. \u676d\u5dde\u7535\u5b50\u79d1\u6280\u5927\u5b66", + "2. \u4e2d\u56fd\u8054\u901a(\u6d59\u6c5f)\u4ea7\u4e1a\u4e92\u8054\u7f51\u6709\u9650\u516c\u53f8" + ], + "abstract": "SAR\uff08SyntheticApertureRadar\uff0c\u5408\u6210\u5b54\u5f84\u96f7\u8fbe\uff09\u56fe\u50cf\u7684\u5bf9\u6297\u6837\u672c\u751f\u6210\u5728\u5f53\u524d\u5df2\u7ecf\u6709\u5f88\u591a\u65b9\u6cd5\uff0c\u4f46\u4ecd\u5b58\u5728\u7740\u5bf9\u6297\u6837\u672c\u6270\u52a8\u91cf\u8f83\u5927\u3001\u8bad\u7ec3\u4e0d\u7a33\u5b9a\u4ee5\u53ca\u5bf9\u6297\u6837\u672c\u7684\u8d28\u91cf\u65e0\u6cd5\u4fdd\u8bc1\u7b49\u95ee\u9898\u3002\u9488\u5bf9\u4e0a\u8ff0\u95ee\u9898\uff0c\u63d0\u51fa\u4e86\u4e00\u79cdSAR\u56fe\u50cf\u5bf9\u6297\u6837\u672c\u751f\u6210\u6a21\u578b\uff0c\u8be5\u6a21\u578b\u57fa\u4e8eAdvGAN\u6a21\u578b\u67b6\u6784\uff0c\u9996\u5148\u6839\u636eSAR\u56fe\u50cf\u7684\u7279\u70b9\u8bbe\u8ba1\u4e86\u4e00\u79cd\u7531\u589e\u5f3aLee\u6ee4\u6ce2\u5668\u548cOTSU\uff08\u6700\u5927\u7c7b\u95f4\u65b9\u5dee\u6cd5\uff09\u81ea\u9002\u5e94\u9608\u503c\u5206\u5272\u7b49\u6a21\u5757\u7ec4\u6210\u7684\u63a9\u6a21\u63d0\u53d6\u6a21\u5757\uff0c\u8fd9\u79cd\u65b9\u6cd5\u4ea7\u751f\u7684\u6270\u52a8\u91cf\u66f4\u5c0f\uff0c\u4e0e\u539f\u59cb\u6837\u672c\u7684SSIM\uff08Structural Similarity\uff0c\u7ed3\u6784\u76f8\u4f3c\u6027\uff09\u503c\u8fbe\u52300.997\u4ee5\u4e0a\u3002\u5176\u6b21\u5c06\u6539\u8fdb\u7684RaGAN\u635f\u5931\u5f15\u5165\u5230AdvGAN\u4e2d\uff0c\u4f7f\u7528\u76f8\u5bf9\u5747\u503c\u5224\u522b\u5668\uff0c\u8ba9\u5224\u522b\u5668\u5728\u8bad\u7ec3\u65f6\u540c\u65f6\u4f9d\u8d56\u4e8e\u771f\u5b9e\u6570\u636e\u548c\u751f\u6210\u7684\u6570\u636e\uff0c\u63d0\u9ad8\u4e86\u8bad\u7ec3\u7684\u7a33\u5b9a\u6027\u4e0e\u653b\u51fb\u6548\u679c\u3002\u5728MSTAR\u6570\u636e\u96c6\u4e0a\u4e0e\u76f8\u5173\u65b9\u6cd5\u8fdb\u884c\u4e86\u5b9e\u9a8c\u5bf9\u6bd4\uff0c\u5b9e\u9a8c\u8868\u660e\uff0c\u6b64\u65b9\u6cd5\u751f\u6210\u7684SAR\u56fe\u50cf\u5bf9\u6297\u6837\u672c\u5728\u653b\u51fb\u9632\u5fa1\u6a21\u578b\u65f6\u7684\u653b\u51fb\u6210\u529f\u7387\u8f83\u4f20\u7edf\u65b9\u6cd5\u63d0\u9ad8\u4e8610%\uff5e15%\u3002" + }, + { + "name": "\u56fe\u795e\u7ecf\u7f51\u7edc\u5bf9\u6297\u653b\u51fb\u4e0e\u9c81\u68d2\u6027\u8bc4\u6d4b\u524d\u6cbf\u8fdb\u5c55", + "authors": [ + "\u5434\u6d9b1,2,3", + "\u66f9\u65b0\u6c761,2", + "\u5148\u5174\u5e731,2,3", + "\u8881\u97161,2", + "\u5f20\u6b8a3", + "\u5d14\u707f\u4e00\u661f1,2", + "\u7530\u4f833" + ], + "affiliations": [ + "1. \u91cd\u5e86\u90ae\u7535\u5927\u5b66\u7f51\u7edc\u7a7a\u95f4\u5b89\u5168\u4e0e\u4fe1\u606f\u6cd5\u5b66\u9662", + "2. \u91cd\u5e86\u5e02\u7f51\u7edc\u4e0e\u4fe1\u606f\u5b89\u5168\u6280\u672f\u5de5\u7a0b\u5b9e\u9a8c\u5ba4", + "3. \u91cd\u5e86\u90ae\u7535\u5927\u5b66-\u91cd\u5e86\u4e2d\u56fd\u4e09\u5ce1\u535a\u7269\u9986\u667a\u6167\u6587\u535a\u8054\u5408\u5b9e\u9a8c\u5ba4" + ], + "abstract": "\u8fd1\u5e74\u6765\uff0c\u56fe\u795e\u7ecf\u7f51\u7edc\uff08GNNs\uff09\u9010\u6e10\u6210\u4e3a\u4eba\u5de5\u667a\u80fd\u7684\u91cd\u8981\u7814\u7a76\u65b9\u5411\u3002\u7136\u800c\uff0cGNNs\u7684\u5bf9\u6297\u8106\u5f31\u6027\u4f7f\u5176\u5b9e\u9645\u5e94\u7528\u9762\u4e34\u4e25\u5cfb\u6311\u6218\u3002\u4e3a\u4e86\u5168\u9762\u8ba4\u8bc6GNNs\u5bf9\u6297\u653b\u51fb\u4e0e\u9c81\u68d2\u6027\u8bc4\u6d4b\u7684\u7814\u7a76\u5de5\u4f5c\uff0c\u5bf9\u76f8\u5173\u524d\u6cbf\u8fdb\u5c55\u8fdb\u884c\u68b3\u7406\u548c\u5206\u6790\u8ba8\u8bba\u3002\u9996\u5148\uff0c\u4ecb\u7ecdGNNs\u5bf9\u6297\u653b\u51fb\u7684\u7814\u7a76\u80cc\u666f\uff0c\u7ed9\u51faGNNs\u5bf9\u6297\u653b\u51fb\u7684\u5f62\u5f0f\u5316\u5b9a\u4e49\uff0c\u9610\u8ff0GNNs\u5bf9\u6297\u653b\u51fb\u53ca\u9c81\u68d2\u6027\u8bc4\u6d4b\u7684\u7814\u7a76\u6846\u67b6\u548c\u57fa\u672c\u6982\u5ff5\u3002\u7136\u540e\uff0c\u5bf9GNNs\u5bf9\u6297\u653b\u51fb\u9886\u57df\u6240\u63d0\u5177\u4f53\u65b9\u6cd5\u8fdb\u884c\u4e86\u603b\u7ed3\u548c\u68b3\u7406\uff0c\u5e76\u5bf9\u5176\u4e2d\u7684\u524d\u6cbf\u65b9\u6cd5\u4ece\u5bf9\u6297\u653b\u51fb\u7c7b\u578b\u548c\u653b\u51fb\u76ee\u6807\u8303\u56f4\u7684\u89d2\u5ea6\u8fdb\u884c\u8be6\u7ec6\u5206\u7c7b\u9610\u8ff0\uff0c\u5206\u6790\u4e86\u5b83\u4eec\u7684\u5de5\u4f5c\u673a\u5236\u3001\u539f\u7406\u548c\u4f18\u7f3a\u70b9\u3002\u5176\u6b21\uff0c\u8003\u8651\u5230\u57fa\u4e8e\u5bf9\u6297\u653b\u51fb\u7684\u6a21\u578b\u9c81\u68d2\u6027\u8bc4\u6d4b\u4f9d\u8d56\u4e8e\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\u7684\u9009\u62e9\u548c\u5bf9\u6297\u6270\u52a8\u7a0b\u5ea6\uff0c\u53ea\u80fd\u5b9e\u73b0\u95f4\u63a5\u3001\u5c40\u90e8\u7684\u8bc4\u4ef7\uff0c\u96be\u4ee5\u5168\u9762\u53cd\u6620\u6a21\u578b\u9c81\u68d2\u6027\u7684\u672c\u8d28\u7279\u5f81\uff0c\u4ece\u800c\u7740\u91cd\u5bf9\u6a21\u578b\u9c81\u68d2\u6027\u7684\u76f4\u63a5\u8bc4\u6d4b\u6307\u6807\u8fdb\u884c\u4e86\u68b3\u7406\u548c\u5206\u6790\u3002\u5728\u6b64\u57fa\u7840\u4e0a\uff0c\u4e3a\u4e86\u652f\u6491GNNs\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\u548c\u9c81\u68d2\u6027\u6a21\u578b\u7684\u8bbe\u8ba1\u4e0e\u8bc4\u4ef7\uff0c\u901a\u8fc7\u5b9e\u9a8c\u4ece\u6613\u5b9e\u73b0\u7a0b\u5ea6\u3001\u51c6\u786e\u6027\u3001\u6267\u884c\u65f6\u95f4\u7b49\u65b9\u9762\u5bf9\u4ee3\u8868\u6027\u7684GNNs\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\u8fdb\u884c\u4e86\u5bf9\u6bd4\u5206\u6790\u3002\u6700\u540e\uff0c\u5bf9\u5b58\u5728\u7684\u6311\u6218\u548c\u672a\u6765\u7814\u7a76\u65b9\u5411\u8fdb\u884c\u5c55\u671b\u3002\u603b\u4f53\u800c\u8a00\uff0c\u76ee\u524dGNNs\u5bf9\u6297\u9c81\u68d2\u6027\u7814\u7a76\u4ee5\u53cd\u590d\u5b9e\u9a8c\u4e3a\u4e3b\u3001\u7f3a\u4e4f\u5177\u6709\u6307\u5bfc\u6027\u7684\u7406\u8bba\u6846\u67b6\u3002\u5982\u4f55\u4fdd\u969c\u57fa\u4e8eGNNs\u7684\u6df1\u5ea6\u667a\u80fd\u7cfb\u7edf\u7684\u53ef\u4fe1\u6027\uff0c\u4ecd\u9700\u8fdb\u4e00\u6b65\u7cfb\u7edf\u6027\u7684\u57fa\u7840\u7406\u8bba\u7814\u7a76\u3002 " + }, + { + "name": "\u8bca\u65ad\u548c\u63d0\u9ad8\u8fc1\u79fb\u5b66\u4e60\u6a21\u578b\u9c81\u68d2\u6027\u7684\u53ef\u89c6\u5206\u6790\u65b9\u6cd5", + "authors": [ + "\u5218\u771f", + "\u989c\u83c1", + "\u5434\u5146\u56fd", + "\u6797\u83f2", + "\u5434\u5411\u9633" + ], + "affiliations": [ + "\u676d\u5dde\u7535\u5b50\u79d1\u6280\u5927\u5b66\u8ba1\u7b97\u673a\u5b66\u9662" + ], + "abstract": "\u867d\u7136\u8fc1\u79fb\u5b66\u4e60\u53ef\u4ee5\u4f7f\u5f00\u53d1\u4eba\u5458\u6839\u636e\u590d\u6742\u7684\u9884\u8bad\u7ec3\u6a21\u578b(\u6559\u5e08\u6a21\u578b)\u6784\u5efa\u7b26\u5408\u76ee\u6807\u4efb\u52a1\u7684\u81ea\u5b9a\u4e49\u6a21\u578b(\u5b66\u751f\u6a21\u578b)\uff0c \u4f46\u662f\u8fc1\u79fb\u5b66\u4e60\u4e2d\u7684\u5b66\u751f\u6a21\u578b\u53ef\u80fd\u4f1a\u7ee7\u627f\u6559\u5e08\u6a21\u578b\u4e2d\u7684\u7f3a\u9677\uff0c \u800c\u6a21\u578b\u9c81\u68d2\u6027\u662f\u4f5c\u4e3a\u8861\u91cf\u6a21\u578b\u7f3a\u9677\u7ee7\u627f\u7684\u91cd\u8981\u6307\u6807\u4e4b\u4e00. \u5728\u8fc1\u79fb\u5b66\u4e60\u9886\u57df\u4e2d\uff0c \u901a\u5e38\u4f1a\u8fd0\u7528\u7f3a\u9677\u7f13\u89e3\u6216\u5b66\u751f\u6a21\u578b\u548c\u6559\u5e08\u6a21\u578b\u8054\u5408\u8bad\u7ec3\u7684\u65b9\u6cd5\uff0c \u8fbe\u5230\u51cf\u5c11\u7ee7\u627f\u6559\u5e08\u6a21\u578b\u7684\u7f3a\u9677\u77e5\u8bc6\u76ee\u7684. \u56e0\u6b64\uff0c \u8bba\u6587\u63d0\u51fa\u4e00\u79cd\u7528\u4e8e\u63a2\u7d22\u8fc1\u79fb\u5b66\u4e60\u8fc7\u7a0b\u4e2d\u6a21\u578b\u9c81\u68d2\u6027\u53d8\u5316\u60c5\u51b5\u7684\u53ef\u89c6\u5206\u6790\u65b9\u6cd5\uff0c \u5e76\u6784\u5efa\u4e86\u76f8\u5e94\u7684\u539f\u578b\u7cfb\u7edf\u2014\u2014TLMRVis. \u8be5\u65b9\u6cd5\u9996\u5148\u8ba1\u7b97\u4e86\u5b66\u751f\u6a21\u578b\u7684\u9c81\u68d2\u6027\u80fd\u6307\u6807; \u5176\u6b21\u5728\u6570\u636e\u5b9e\u4f8b\u5c42\u9762\u5c55\u793a\u6a21\u578b\u5404\u7c7b\u522b\u7684\u8868\u73b0\u6027\u80fd; \u7136\u540e\u5728\u5b9e\u4f8b\u7279\u5f81\u5c42\u9762\u901a\u8fc7\u6a21\u578b\u62bd\u8c61\u5316\u65b9\u5f0f\u53bb\u63ed\u793a\u6559\u5e08\u6a21\u578b\u548c\u5b66\u751f\u6a21\u578b\u4e4b\u95f4\u7ee7\u627f\u7684\u91cd\u7528\u77e5\u8bc6; \u6700\u540e\u7ed3\u5408\u6a21\u578b\u5207\u7247\u65b9\u6cd5\u6539\u5584\u6a21\u578b\u7684\u7f3a\u9677\u7ee7\u627f\u7528\u4ee5\u63d0\u9ad8\u6a21\u578b\u9c81\u68d2\u6027. \u540c\u65f6\uff0c TLMRVis\u7cfb\u7edf\u4e0d\u4ec5\u7ed3\u5408\u591a\u79cd\u53ef\u89c6\u5316\u65b9\u6cd5\u5c55\u793a\u591a\u79cd\u5b66\u751f\u6a21\u578b\u548c\u6559\u5e08\u6a21\u578b\u4e4b\u95f4\u7684\u5f02\u540c\u70b9\uff0c \u800c\u4e14\u901a\u8fc7\u5f15\u5165\u7f3a\u9677\u7f13\u89e3\u6280\u672f\u6765\u67e5\u770b\u548c\u8bca\u65ad\u6559\u5e08\u6a21\u578b\u548c\u5b66\u751f\u6a21\u578b\u7684\u6027\u80fd\u53d8\u5316\u548c\u5e95\u5c42\u9884\u6d4b\u884c\u4e3a\u673a\u5236. 2\u4e2a\u6848\u4f8b\u7684\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c TLMRVis\u7cfb\u7edf\u53ef\u4ee5\u5e2e\u52a9\u7528\u6237\u5206\u6790\u8fc1\u79fb\u5b66\u4e60\u4e2d\u6a21\u578b\u7684\u9c81\u68d2\u6027\u3001\u6a21\u578b\u7ee7\u627f\u7684\u7f3a\u9677\u77e5\u8bc6\u548c\u6a21\u578b\u7f3a\u9677\u6539\u5584\u540e\u7684\u6027\u80fd\u53d8\u5316." + }, + { + "name": "\u57fa\u4e8e\u8fd1\u7aef\u7ebf\u6027\u7ec4\u5408\u7684\u4fe1\u53f7\u8bc6\u522b\u795e\u7ecf\u7f51\u7edc\u9ed1\u76d2\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5", + "authors": [ + "\u90ed\u5b87\u7426", + "\u674e\u4e1c\u9633", + "\u95eb\u9554", + "\u738b\u6797\u5143" + ], + "affiliations": [ + "\u6218\u7565\u652f\u63f4\u90e8\u961f\u4fe1\u606f\u5de5\u7a0b\u5927\u5b66\u6210\u50cf\u4e0e\u667a\u80fd\u5904\u7406\u5b9e\u9a8c\u5ba4" + ], + "abstract": "\u968f\u7740\u6df1\u5ea6\u5b66\u4e60\u5728\u65e0\u7ebf\u901a\u4fe1\u9886\u57df\u7279\u522b\u662f\u4fe1\u53f7\u8c03\u5236\u8bc6\u522b\u65b9\u5411\u7684\u5e7f\u6cdb\u5e94\u7528\uff0c\u795e\u7ecf\u7f51\u7edc\u6613\u53d7\u5bf9\u6297\u6837\u672c\u653b\u51fb\u7684\u95ee\u9898\u540c\u6837\u5f71\u54cd\u7740\u65e0\u7ebf\u901a\u4fe1\u7684\u5b89\u5168\u3002\u9488\u5bf9\u65e0\u7ebf\u4fe1\u53f7\u5728\u901a\u4fe1\u4e2d\u96be\u4ee5\u5b9e\u65f6\u83b7\u5f97\u795e\u7ecf\u7f51\u7edc\u53cd\u9988\u4e14\u53ea\u80fd\u8bbf\u95ee\u8bc6\u522b\u7ed3\u679c\u7684\u9ed1\u76d2\u653b\u51fb\u573a\u666f\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u8fd1\u7aef\u7ebf\u6027\u7ec4\u5408\u7684\u9ed1\u76d2\u67e5\u8be2\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\u3002\u8be5\u65b9\u6cd5\u9996\u5148\u5728\u6570\u636e\u96c6\u7684\u4e00\u4e2a\u5b50\u96c6\u4e0a\uff0c\u5bf9\u6bcf\u4e2a\u539f\u59cb\u4fe1\u53f7\u6837\u672c\u8fdb\u884c\u8fd1\u7aef\u7ebf\u6027\u7ec4\u5408\uff0c\u5373\u5728\u975e\u5e38\u9760\u8fd1\u539f\u59cb\u4fe1\u53f7\u7684\u8303\u56f4\u5185\u4e0e\u76ee\u6807\u4fe1\u53f7\u8fdb\u884c\u7ebf\u6027\u7ec4\u5408\uff08\u52a0\u6743\u7cfb\u6570\u4e0d\u5927\u4e8e0.05\uff09\uff0c\u5e76\u5c06\u5176\u8f93\u5165\u5f85\u653b\u51fb\u7f51\u7edc\u67e5\u8be2\u8bc6\u522b\u7ed3\u679c\u3002\u901a\u8fc7\u7edf\u8ba1\u7f51\u7edc\u5bf9\u5168\u90e8\u8fd1\u7aef\u7ebf\u6027\u7ec4\u5408\u8bc6\u522b\u51fa\u9519\u7684\u6570\u91cf\uff0c\u786e\u5b9a\u6bcf\u7c7b\u539f\u59cb\u4fe1\u53f7\u6700\u5bb9\u6613\u53d7\u5230\u7ebf\u6027\u7ec4\u5408\u5f71\u54cd\u7684\u7279\u5b9a\u76ee\u6807\u4fe1\u53f7\uff0c\u5c06\u5176\u79f0\u4e3a\u6700\u4f73\u6270\u52a8\u4fe1\u53f7\u3002\u5728\u653b\u51fb\u6d4b\u8bd5\u65f6\uff0c\u6839\u636e\u4fe1\u53f7\u7684\u7c7b\u522b\u9009\u62e9\u5bf9\u5e94\u6700\u4f73\u6270\u52a8\u4fe1\u53f7\u6267\u884c\u8fd1\u7aef\u7ebf\u6027\u7ec4\u5408\uff0c\u751f\u6210\u5bf9\u6297\u6837\u672c\u3002\u5b9e\u9a8c\u7ed3\u679c\u663e\u793a\uff0c\u8be5\u65b9\u6cd5\u5728\u9009\u5b9a\u5b50\u96c6\u4e0a\u9488\u5bf9\u6bcf\u79cd\u8c03\u5236\u7c7b\u522b\u7684\u6700\u4f73\u6270\u52a8\u4fe1\u53f7\uff0c\u6dfb\u52a0\u5728\u5168\u90e8\u6570\u636e\u96c6\u4e0a\u80fd\u5c06\u795e\u7ecf\u7f51\u7edc\u8bc6\u522b\u51c6\u786e\u7387\u4ece94%\u964d\u523050%\uff0c\u4e14\u76f8\u8f83\u4e8e\u6dfb\u52a0\u968f\u673a\u566a\u58f0\u653b\u51fb\u7684\u6270\u52a8\u529f\u7387\u66f4\u5c0f\u3002\u6b64\u5916\uff0c\u751f\u6210\u7684\u5bf9\u6297\u6837\u672c\u5bf9\u4e8e\u7ed3\u6784\u8fd1\u4f3c\u7684\u795e\u7ecf\u7f51\u7edc\u5177\u6709\u4e00\u5b9a\u8fc1\u79fb\u6027\u3002\u8fd9\u79cd\u65b9\u6cd5\u5728\u7edf\u8ba1\u67e5\u8be2\u540e\u751f\u6210\u65b0\u7684\u5bf9\u6297\u6837\u672c\u65f6\uff0c\u6613\u4e8e\u5b9e\u73b0\u4e14\u65e0\u9700\u518d\u8fdb\u884c\u9ed1\u76d2\u67e5\u8be2\u3002" + }, + { + "name": "\u57fa\u4e8e\u751f\u6210\u5bf9\u6297\u7f51\u7edc\u7684\u6df1\u5ea6\u4f2a\u9020\u8de8\u6a21\u578b\u9632\u5fa1\u65b9\u6cd5", + "authors": [ + "\u6234\u78ca", + "\u66f9\u6797", + "\u90ed\u4e9a\u7537", + "\u5f20\u5e06", + "\u675c\u5eb7\u5b81" + ], + "affiliations": [ + "\u5317\u4eac\u4fe1\u606f\u79d1\u6280\u5927\u5b66\u4fe1\u606f\u4e0e\u901a\u4fe1\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u4e3a\u4e86\u964d\u4f4e\u6df1\u5ea6\u4f2a\u9020\uff08deepfake\uff09\u6280\u672f\u6ee5\u7528\u5e26\u6765\u7684\u793e\u4f1a\u98ce\u9669\uff0c\u63d0\u51fa\u4e00\u79cd\u57fa\u4e8e\u751f\u6210\u5bf9\u6297\u7f51\u7edc\u7684\u4e3b\u52a8\u9632\u5fa1\u6df1\u5ea6\u4f2a\u9020\u65b9\u6cd5\uff0c\u901a\u8fc7\u5728\u539f\u59cb\u56fe\u50cf\u4e0a\u589e\u52a0\u5fae\u5f31\u6270\u52a8\u5236\u4f5c\u5bf9\u6297\u6837\u672c\uff0c\u4f7f\u591a\u4e2a\u4f2a\u9020\u6a21\u578b\u8f93\u51fa\u4ea7\u751f\u660e\u663e\u5931\u771f\u3002\u63d0\u51fa\u7684\u6a21\u578b\u7531\u5bf9\u6297\u6837\u672c\u751f\u6210\u6a21\u5757\u548c\u5bf9\u6297\u6837\u672c\u4f18\u5316\u6a21\u5757\u7ec4\u6210\u3002\u5bf9\u6297\u6837\u672c\u751f\u6210\u6a21\u5757\u5305\u62ec\u751f\u6210\u5668\u548c\u9274\u522b\u5668\uff0c\u751f\u6210\u5668\u63a5\u6536\u539f\u59cb\u56fe\u50cf\u751f\u6210\u6270\u52a8\u540e\uff0c\u901a\u8fc7\u5bf9\u6297\u8bad\u7ec3\u7ea6\u675f\u6270\u52a8\u7684\u7a7a\u95f4\u5206\u5e03\uff0c\u964d\u4f4e\u6270\u52a8\u7684\u89c6\u89c9\u611f\u77e5\uff0c\u63d0\u9ad8\u5bf9\u6297\u6837\u672c\u7684\u771f\u5b9e\u6027\uff1b\u5bf9\u6297\u6837\u672c\u4f18\u5316\u6a21\u5757\u7531\u57fa\u7840\u5bf9\u6297\u6c34\u5370\u3001\u6df1\u5ea6\u4f2a\u9020\u6a21\u578b\u548c\u9274\u522b\u5668\u7b49\u7ec4\u6210\uff0c\u901a\u8fc7\u6a21\u62df\u9ed1\u76d2\u573a\u666f\u4e0b\u653b\u51fb\u591a\u4e2a\u6df1\u5ea6\u4f2a\u9020\u6a21\u578b\uff0c\u63d0\u9ad8\u5bf9\u6297\u6837\u672c\u7684\u653b\u51fb\u6027\u548c\u8fc1\u79fb\u6027\u3002\u5728\u5e38\u7528\u6df1\u5ea6\u4f2a\u9020\u6570\u636e\u96c6CelebA\u548cLFW\u8fdb\u884c\u4e86\u8bad\u7ec3\u548c\u6d4b\u8bd5\uff0c\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u76f8\u6bd4\u73b0\u6709\u4e3b\u52a8\u9632\u5fa1\u65b9\u6cd5\uff0c\u672c\u6587\u5728\u5b9e\u73b0\u8de8\u6a21\u578b\u4e3b\u52a8\u9632\u5fa1\u7684\u57fa\u7840\u4e0a\uff0c\u9632\u5fa1\u6210\u529f\u7387\u8fbe\u523085%\u4ee5\u4e0a\uff0c\u5e76\u4e14\u5bf9\u6297\u6837\u672c\u751f\u6210\u6548\u7387\u6bd4\u4f20\u7edf\u7b97\u6cd5\u63d0\u9ad820\uff5e30\u500d\u3002" + }, + { + "name": "\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u5bf9\u6297\u6837\u672c\u653b\u51fb\u7814\u7a76", + "authors": [ + "\u5f20\u5b66\u519b1", + "\u5e2d\u963f\u53cb1", + "\u52a0\u5c0f\u7ea21", + "\u5f20\u658c1", + "\u674e\u68851", + "\u675c\u6653\u521a2", + "\u9ec4\u6d77\u71d51" + ], + "affiliations": [ + "1. \u5170\u5dde\u4ea4\u901a\u5927\u5b66\u7535\u5b50\u4e0e\u4fe1\u606f\u5de5\u7a0b\u5b66\u9662", + "2. \u9655\u897f\u79d1\u6280\u5927\u5b66\u7535\u5b50\u4fe1\u606f\u4e0e\u4eba\u5de5\u667a\u80fd\u5b66\u9662" + ], + "abstract": "\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u7cfb\u7edf\u56e0\u5176\u80fd\u591f\u6709\u6548\u62bd\u53d6RSS\u6307\u7eb9\u6570\u636e\u7684\u6df1\u5c42\u7279\u5f81\u800c\u5927\u5e45\u63d0\u9ad8\u4e86\u5ba4\u5185\u5b9a\u4f4d\u6027\u80fd\uff0c\u4f46\u8be5\u7c7b\u65b9\u6cd5\u9700\u8981\u5927\u91cf\u3001\u591a\u6837\u5316\u7684RSS\u6307\u7eb9\u6570\u636e\u8bad\u7ec3\u6a21\u578b\uff0c\u5e76\u4e14\u5bf9\u5176\u5b89\u5168\u6f0f\u6d1e\u4e5f\u7f3a\u4e4f\u5145\u5206\u7684\u7814\u7a76\uff0c\u8fd9\u4e9b\u5b89\u5168\u6f0f\u6d1e\u6e90\u4e8e\u65e0\u7ebfWi-Fi\u5a92\u4f53\u7684\u5f00\u653e\u6027\u548c\u5206\u7c7b\u5668\u7684\u56fa\u6709\u7f3a\u9677\uff08\u5982\u6613\u906d\u53d7\u5bf9\u6297\u6027\u653b\u51fb\u7b49\uff09\u3002\u4e3a\u6b64\uff0c\u5bf9\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684RSS\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u7cfb\u7edf\u7684\u5bf9\u6297\u6027\u653b\u51fb\u8fdb\u884c\u7814\u7a76\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8eWi-Fi\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u7684\u5bf9\u6297\u6837\u672c\u653b\u51fb\u6846\u67b6\uff0c\u5e76\u5229\u7528\u8be5\u6846\u67b6\u7814\u7a76\u4e86\u5bf9\u6297\u653b\u51fb\u5bf9\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684RSS\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u6a21\u578b\u6027\u80fd\u7684\u5f71\u54cd\u3002\u8be5\u6846\u67b6\u5305\u542b\u79bb\u7ebf\u8bad\u7ec3\u548c\u5728\u7ebf\u5b9a\u4f4d\u4e24\u4e2a\u9636\u6bb5\uff0c\u5728\u79bb\u7ebf\u8bad\u7ec3\u9636\u6bb5\uff0c\u8bbe\u8ba1\u9002\u7528\u4e8e\u589e\u5e7fWi-Fi RSS\u6307\u7eb9\u6570\u636e\u7684\u6761\u4ef6\u751f\u6210\u5bf9\u6297\u7f51\u7edc\uff08CGAN\uff09\u6765\u751f\u6210\u5927\u91cf\u3001\u591a\u6837\u5316\u7684RSS\u6307\u7eb9\u6570\u636e\u8bad\u7ec3\u9ad8\u9c81\u68d2\u7684\u5ba4\u5185\u5b9a\u4f4d\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\uff1b\u5728\u7ebf\u5b9a\u4f4d\u9636\u6bb5\uff0c\u6784\u9020\u6700\u5f3a\u7684\u4e00\u9636\u653b\u51fb\u7b56\u7565\u6765\u751f\u6210\u9488\u5bf9Wi-Fi RSS\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u7cfb\u7edf\u7684\u6709\u6548RSS\u5bf9\u6297\u6837\u672c\uff0c\u7814\u7a76\u5bf9\u6297\u653b\u51fb\u5bf9\u4e0d\u540c\u5ba4\u5185\u5b9a\u4f4d\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u6027\u80fd\u7684\u5f71\u54cd\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff1a\u5728\u516c\u5f00UJIIndoorLoc\u6570\u636e\u96c6\u4e0a\uff0c\u7531\u6240\u63d0\u6846\u67b6\u751f\u6210\u7684RSS\u6307\u7eb9\u5bf9\u6297\u6837\u672c\u5bf9\u73b0\u6709\u7684CNN\u3001DNN\u3001MLP\u3001pixeldp\uff3fCNN\u7684\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u6a21\u578b\u7684\u653b\u51fb\u6210\u529f\u7387\u5206\u522b\u8fbe\u523094.1%\uff0c63.75%\uff0c43.45%\uff0c72.5%\uff1b\u800c\u4e14\uff0c\u5bf9\u7531CGAN\u7f51\u7edc\u589e\u5e7f\u6570\u636e\u8bad\u7ec3\u7684\u4e0a\u8ff0\u56db\u79cd\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u6a21\u578b\u7684\u653b\u51fb\u6210\u529f\u7387\u4ecd\u5206\u522b\u8fbe\u5230\u4e8684.95%\uff0c44.8%\uff0c15.7%\uff0c11.5%\uff1b\u56e0\u6b64\uff0c\u73b0\u6709\u7684\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u6307\u7eb9\u5ba4\u5185\u5b9a\u4f4d\u6a21\u578b\u6613\u906d\u53d7\u5bf9\u6297\u6837\u672c\u653b\u51fb\u7684\u5f71\u54cd\uff0c\u7531\u771f\u5b9e\u6570\u636e\u548c\u589e\u5e7f\u6570\u636e\u6df7\u5408\u8bad\u7ec3\u7684\u5ba4\u5185\u5b9a\u4f4d\u6a21\u578b\u5728\u9762\u4e34\u5bf9\u6297\u6837\u672c\u653b\u51fb\u65f6\u5177\u6709\u66f4\u597d\u7684\u9c81\u68d2\u6027\u3002 " + }, + { + "name": "\u57fa\u4e8eGAN\u7684\u65e0\u6570\u636e\u9ed1\u76d2\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5", + "authors": [ + "\u8d75\u6069\u6d69", + "\u51cc\u6377" + ], + "affiliations": [ + "\u5e7f\u4e1c\u5de5\u4e1a\u5927\u5b66\u8ba1\u7b97\u673a\u5b66\u9662" + ], + "abstract": "\u5bf9\u6297\u6837\u672c\u80fd\u591f\u4f7f\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u4ee5\u9ad8\u7f6e\u4fe1\u5ea6\u8f93\u51fa\u9519\u8bef\u7684\u7ed3\u679c\u3002\u5728\u9ed1\u76d2\u653b\u51fb\u4e2d\uff0c\u73b0\u6709\u7684\u66ff\u4ee3\u6a21\u578b\u8bad\u7ec3\u65b9\u6cd5\u9700\u8981\u76ee\u6807\u6a21\u578b\u5168\u90e8\u6216\u90e8\u5206\u8bad\u7ec3\u6570\u636e\u624d\u80fd\u53d6\u5f97\u8f83\u597d\u7684\u653b\u51fb\u6548\u679c\uff0c\u4f46\u5b9e\u9645\u5e94\u7528\u4e2d\u76ee\u6807\u6a21\u578b\u7684\u8bad\u7ec3\u6570\u636e\u96be\u4ee5\u83b7\u53d6\u3002\u56e0\u6b64\uff0c\u6587\u4e2d\u63d0\u51fa\u4e00\u79cd\u57fa\u4e8eGAN\u7684\u65e0\u6570\u636e\u9ed1\u76d2\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\u3002\u65e0\u9700\u76ee\u6807\u6a21\u578b\u7684\u8bad\u7ec3\u6570\u636e\uff0c\u4f7f\u7528\u6df7\u5408\u6807\u7b7e\u4fe1\u606f\u7684\u566a\u58f0\u751f\u6210\u66ff\u4ee3\u6a21\u578b\u6240\u9700\u7684\u8bad\u7ec3\u6837\u672c\uff0c\u901a\u8fc7\u76ee\u6807\u6a21\u578b\u7684\u6807\u8bb0\u4fe1\u606f\u4ee5\u53ca\u591a\u6837\u5316\u635f\u5931\u51fd\u6570\u4f7f\u8bad\u7ec3\u6837\u672c\u5206\u5e03\u5747\u5300\u4e14\u5305\u542b\u66f4\u591a\u7279\u5f81\u4fe1\u606f\uff0c\u8fdb\u800c\u4f7f\u66ff\u4ee3\u6a21\u578b\u9ad8\u6548\u5b66\u4e60\u76ee\u6807\u6a21\u578b\u7684\u5206\u7c7b\u529f\u80fd\u3002\u5bf9\u6bd4DaST\u548cMAZE\uff0c\u6587\u4e2d\u65b9\u6cd5\u5728\u964d\u4f4e35%-60%\u7684\u5bf9\u6297\u6270\u52a8\u548c\u67e5\u8be2\u6b21\u6570\u7684\u540c\u65f6\u5bf9CIFAR-100\u3001CIFAR-10\u3001SVHN\u3001FMNIST\u3001MNIST\u4e94\u4e2a\u6570\u636e\u96c6\u7684FGSM\u3001BIM\u3001PGD\u4e09\u79cd\u653b\u51fb\u7684\u6210\u529f\u7387\u5e73\u5747\u63d0\u9ad86%-10%\u3002\u5e76\u4e14\u5728\u5b9e\u9645\u5e94\u7528\u4e2d\u7684\u9ed1\u76d2\u6a21\u578b\u573a\u666fMicrosoft Azure\u53d6\u5f9778%\u4ee5\u4e0a\u7684\u653b\u51fb\u6210\u529f\u7387\u3002" + }, + { + "name": "\u9762\u5411\u9c81\u68d2\u56fe\u7ed3\u6784\u9632\u5fa1\u7684\u8fc7\u53c2\u6570\u5316\u56fe\u795e\u7ecf\u7f51\u7edc", + "authors": [ + "\u521d\u65ed1", + "\u9a6c\u8f9b\u5b872,3", + "\u6797\u96332,3", + "\u738b\u946b1,4", + "\u738b\u4e9a\u6c993,5", + "\u6731\u6587\u6b661,4", + "\u6885\u5b8f3" + ], + "affiliations": [ + "1. \u6e05\u534e\u5927\u5b66\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u7cfb", + "2. \u5317\u4eac\u5927\u5b66\u8ba1\u7b97\u673a\u5b66\u9662", + "3. \u9ad8\u53ef\u4fe1\u8f6f\u4ef6\u6280\u672f\u6559\u80b2\u90e8\u91cd\u70b9\u5b9e\u9a8c\u5ba4(\u5317\u4eac\u5927\u5b66)", + "4. \u6e05\u534e\u5927\u5b66\u5317\u4eac\u4fe1\u606f\u79d1\u5b66\u4e0e\u6280\u672f\u56fd\u5bb6\u7814\u7a76\u4e2d\u5fc3", + "5. \u5317\u4eac\u5927\u5b66\u8f6f\u4ef6\u5de5\u7a0b\u56fd\u5bb6\u5de5\u7a0b\u4e2d\u5fc3" + ], + "abstract": "\u56fe\u6570\u636e\u5728\u73b0\u5b9e\u5e94\u7528\u4e2d\u666e\u904d\u5b58\u5728,\u56fe\u795e\u7ecf\u7f51\u7edc(GNN)\u88ab\u5e7f\u6cdb\u5e94\u7528\u4e8e\u5206\u6790\u56fe\u6570\u636e,\u7136\u800cGNN\u7684\u6027\u80fd\u4f1a\u88ab\u56fe\u7ed3\u6784\u4e0a\u7684\u5bf9\u6297\u653b\u51fb\u5267\u70c8\u5f71\u54cd.\u5e94\u5bf9\u56fe\u7ed3\u6784\u4e0a\u7684\u5bf9\u6297\u653b\u51fb,\u73b0\u6709\u7684\u9632\u5fa1\u65b9\u6cd5\u4e00\u822c\u57fa\u4e8e\u56fe\u5185\u805a\u5148\u9a8c\u8fdb\u884c\u4f4e\u79e9\u56fe\u7ed3\u6784\u91cd\u6784.\u4f46\u662f\u73b0\u6709\u7684\u56fe\u7ed3\u6784\u5bf9\u6297\u9632\u5fa1\u65b9\u6cd5\u65e0\u6cd5\u81ea\u9002\u5e94\u79e9\u771f\u503c\u8fdb\u884c\u4f4e\u79e9\u56fe\u7ed3\u6784\u91cd\u6784,\u540c\u65f6\u4f4e\u79e9\u56fe\u7ed3\u6784\u4e0e\u4e0b\u6e38\u4efb\u52a1\u8bed\u4e49\u5b58\u5728\u9519\u914d.\u4e3a\u4e86\u89e3\u51b3\u4ee5\u4e0a\u95ee\u9898,\u57fa\u4e8e\u8fc7\u53c2\u6570\u5316\u7684\u9690\u5f0f\u6b63\u5219\u6548\u5e94\u63d0\u51fa\u8fc7\u53c2\u6570\u5316\u56fe\u795e\u7ecf\u7f51\u7edc(OPGNN)\u65b9\u6cd5,\u5e76\u5f62\u5f0f\u5316\u8bc1\u660e\u6240\u63d0\u65b9\u6cd5\u53ef\u4ee5\u81ea\u9002\u5e94\u6c42\u89e3\u4f4e\u79e9\u56fe\u7ed3\u6784,\u540c\u65f6\u8bc1\u660e\u8282\u70b9\u6df1\u5c42\u8868\u5f81\u4e0a\u7684\u8fc7\u53c2\u6570\u5316\u6b8b\u5dee\u94fe\u63a5\u53ef\u4ee5\u6709\u6548\u89e3\u51b3\u8bed\u4e49\u9519\u914d.\u5728\u771f\u5b9e\u6570\u636e\u96c6\u4e0a\u7684\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e, OPGNN\u65b9\u6cd5\u76f8\u5bf9\u4e8e\u73b0\u6709\u57fa\u7ebf\u65b9\u6cd5\u5177\u6709\u66f4\u597d\u7684\u9c81\u68d2\u6027,\u540c\u65f6, OPGNN\u65b9\u6cd5\u6846\u67b6\u5728\u4e0d\u540c\u7684\u56fe\u795e\u7ecf\u7f51\u7edc\u9aa8\u5e72\u4e0a\u5982GCN\u3001APPNP\u548cGPRGNN\u4e0a\u663e\u8457\u6709\u6548." + }, + { + "name": "\u57fa\u4e8e\u751f\u6210\u5f0f\u81ea\u76d1\u7763\u5b66\u4e60\u7684\u5bf9\u6297\u6837\u672c\u5206\u7c7b\u7b97\u6cd5", + "authors": [ + "\u9633\u5e061", + "\u9b4f\u5baa2,3", + "\u90ed\u6770\u9f992,3", + "\u90d1\u5efa\u6f332,3", + "\u5170\u6d772" + ], + "affiliations": [ + "1. \u798f\u5dde\u5927\u5b66\u5148\u8fdb\u5236\u9020\u5b66\u9662", + "2. \u4e2d\u56fd\u79d1\u5b66\u9662\u798f\u5efa\u7269\u8d28\u7ed3\u6784\u7814\u7a76\u6240", + "3. \u4e2d\u56fd\u79d1\u5b66\u9662\u6d77\u897f\u7814\u7a76\u9662\u6cc9\u5dde\u88c5\u5907\u5236\u9020\u7814\u7a76\u4e2d\u5fc3" + ], + "abstract": "\u5bf9\u6297\u6837\u672c\u5e38\u5e38\u88ab\u89c6\u4e3a\u5bf9\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u9c81\u68d2\u6027\u7684\u5a01\u80c1\uff0c\u800c\u73b0\u6709\u5bf9\u6297\u8bad\u7ec3\u5f80\u5f80\u4f1a\u964d\u4f4e\u5206\u7c7b\u7f51\u7edc\u7684\u6cdb\u5316\u7cbe\u5ea6\uff0c\u5bfc\u81f4\u5176\u5bf9\u539f\u59cb\u6837\u672c\u7684\u5206\u7c7b\u6548\u679c\u964d\u4f4e\u3002\u56e0\u6b64\uff0c\u63d0\u51fa\u4e86\u4e00\u4e2a\u57fa\u4e8e\u751f\u6210\u5f0f\u81ea\u76d1\u7763\u5b66\u4e60\u7684\u5bf9\u6297\u6837\u672c\u5206\u7c7b\u7b97\u6cd5\uff0c\u901a\u8fc7\u81ea\u76d1\u7763\u5b66\u4e60\u8bad\u7ec3\u751f\u6210\u5f0f\u6a21\u578b\u83b7\u53d6\u56fe\u50cf\u6570\u636e\u6f5c\u5728\u7279\u5f81\u7684\u80fd\u529b\uff0c\u5e76\u57fa\u4e8e\u8be5\u6a21\u578b\u5b9e\u73b0\u5bf9\u6297\u6837\u672c\u7684\u7279\u5f81\u7b5b\u9009\uff0c\u800c\u540e\u5c06\u5176\u4e2d\u6709\u76ca\u5206\u7c7b\u7684\u4fe1\u606f\u53cd\u9988\u7ed9\u5206\u7c7b\u6a21\u578b\u3002\u6700\u540e\u8fdb\u884c\u8054\u5408\u5b66\u4e60\uff0c\u5b8c\u6210\u7aef\u5230\u7aef\u7684\u5168\u5c40\u8bad\u7ec3\uff0c\u8fdb\u4e00\u6b65\u5b9e\u73b0\u5206\u7c7b\u6a21\u578b\u6cdb\u5316\u7cbe\u5ea6\u7684\u63d0\u5347\u3002\u5728MNIST\u3001CIFAR10\u548cCIFAR100\u6570\u636e\u96c6\u4e0a\u7684\u5b9e\u9a8c\u7ed3\u679c\u663e\u793a\uff0c\u4e0e\u6807\u51c6\u8bad\u7ec3\u76f8\u6bd4\uff0c\u8be5\u7b97\u6cd5\u5c06\u5206\u7c7b\u7cbe\u5ea6\u5206\u522b\u63d0\u9ad8\u4e860.06%\u30011.34%\u30010.89%\uff0c\u8fbe\u523099.70%\u300184.34%\u300163.65%\u3002\u7ed3\u679c\u8bc1\u660e\uff0c\u8be5\u7b97\u6cd5\u514b\u670d\u4e86\u4f20\u7edf\u5bf9\u6297\u8bad\u7ec3\u964d\u4f4e\u6a21\u578b\u6cdb\u5316\u6027\u80fd\u7684\u56fa\u6709\u7f3a\u70b9\uff0c\u5e76\u8fdb\u4e00\u6b65\u63d0\u9ad8\u4e86\u5206\u7c7b\u7f51\u7edc\u7684\u7cbe\u5ea6\u3002" + }, + { + "name": "\u65f6\u9891\u5206\u533a\u6270\u52a8\u5b9e\u73b0\u97f3\u9891\u5206\u7c7b\u5bf9\u6297\u6837\u672c\u751f\u6210", + "authors": [ + "\u5f20\u96c4\u4f1f", + "\u5f20\u5f3a", + "\u6768\u5409\u658c", + "\u5b59\u8499", + "\u674e\u6bc5\u8c6a" + ], + "affiliations": [ + "\u9646\u519b\u5de5\u7a0b\u5927\u5b66\u6307\u6325\u63a7\u5236\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u73b0\u6709\u65b9\u6cd5\u751f\u6210\u7684\u97f3\u9891\u5206\u7c7b\u5bf9\u6297\u6837\u672c(adversarial example, AE)\u653b\u51fb\u6210\u529f\u7387\u4f4e\uff0c\u6613\u88ab\u611f\u77e5\u3002\u9274\u4e8e\u6b64\uff0c\u8bbe\u8ba1\u4e86\u4e00\u79cd\u57fa\u4e8e\u65f6\u9891\u5206\u533a\u6270\u52a8(time-frequency partitioned perturbation, TFPP)\u7684\u97f3\u9891AE\u751f\u6210\u6846\u67b6\u3002\u97f3\u9891\u4fe1\u53f7\u7684\u5e45\u5ea6\u8c31\u6839\u636e\u65f6\u9891\u7279\u6027\u88ab\u5212\u5206\u4e3a\u5173\u952e\u548c\u975e\u5173\u952e\u533a\u57df\uff0c\u5e76\u751f\u6210\u76f8\u5e94\u7684\u5bf9\u6297\u6270\u52a8\u3002\u5728TFPP\u57fa\u7840\u4e0a\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u751f\u6210\u5bf9\u6297\u7f51\u7edc(generative adversarial network, GAN)\u7684AE\u751f\u6210\u65b9\u6cd5TFPPGAN,\u4ee5\u5206\u533a\u5e45\u5ea6\u8c31\u4e3a\u8f93\u5165\uff0c\u901a\u8fc7\u5bf9\u6297\u8bad\u7ec3\u81ea\u9002\u5e94\u8c03\u6574\u6270\u52a8\u7ea6\u675f\u7cfb\u6570\uff0c\u540c\u65f6\u4f18\u5316\u5173\u952e\u548c\u975e\u5173\u952e\u533a\u57df\u7684\u6270\u52a8\u30023\u4e2a\u5178\u578b\u97f3\u9891\u5206\u7c7b\u6570\u636e\u96c6\u4e0a\u7684\u5b9e\u9a8c\u8868\u660e\uff0c\u4e0e\u57fa\u7ebf\u65b9\u6cd5\u76f8\u6bd4\uff0cTFPPGAN\u53ef\u5c06AE\u7684\u653b\u51fb\u6210\u529f\u7387\u3001\u4fe1\u566a\u6bd4\u5206\u522b\u63d0\u9ad84.7%\u548c5.5 dB,\u5c06\u751f\u6210\u7684\u8bed\u97f3\u5bf9\u6297\u6837\u672c\u7684\u8d28\u91cf\u611f\u77e5\u8bc4\u4ef7\u5f97\u5206\u63d0\u9ad80.15\u3002\u6b64\u5916\uff0c\u7406\u8bba\u5206\u6790\u4e86TFPP\u6846\u67b6\u4e0e\u5176\u4ed6\u653b\u51fb\u65b9\u6cd5\u76f8\u7ed3\u5408\u7684\u53ef\u884c\u6027\uff0c\u5e76\u901a\u8fc7\u5b9e\u9a8c\u9a8c\u8bc1\u4e86\u8fd9\u79cd\u7ed3\u5408\u7684\u6709\u6548\u6027\u3002" + }, + { + "name": "\u57fa\u4e8e\u635f\u5931\u5e73\u6ed1\u7684\u5bf9\u6297\u6837\u672c\u653b\u51fb\u65b9\u6cd5", + "authors": [ + "\u9ece\u59b9\u7ea21,2", + "\u91d1\u53cc1,2", + "\u675c\u66541,2" + ], + "affiliations": [ + "1. \u5317\u4eac\u4ea4\u901a\u5927\u5b66\u667a\u80fd\u4ea4\u901a\u6570\u636e\u5b89\u5168\u4e0e\u9690\u79c1\u4fdd\u62a4\u6280\u672f\u5317\u4eac\u5e02\u91cd\u70b9\u5b9e\u9a8c\u5ba4", + "2. \u5317\u4eac\u4ea4\u901a\u5927\u5b66\u8ba1\u7b97\u673a\u4e0e\u4fe1\u606f\u6280\u672f\u5b66\u9662" + ], + "abstract": "\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc(DNNs)\u5bb9\u6613\u53d7\u5230\u5bf9\u6297\u6837\u672c\u7684\u653b\u51fb\uff0c\u73b0\u6709\u57fa\u4e8e\u52a8\u91cf\u7684\u5bf9\u6297\u6837\u672c\u751f\u6210\u65b9\u6cd5\u867d\u7136\u53ef\u4ee5\u8fbe\u5230\u63a5\u8fd1100%\u7684\u767d\u76d2\u653b\u51fb\u6210\u529f\u7387\uff0c\u4f46\u662f\u5728\u653b\u51fb\u5176\u4ed6\u6a21\u578b\u65f6\u6548\u679c\u4ecd\u4e0d\u7406\u60f3\uff0c\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387\u8f83\u4f4e\u3002\u9488\u5bf9\u6b64\uff0c\u63d0\u51fa\u4e00\u79cd\u57fa\u4e8e\u635f\u5931\u5e73\u6ed1\u7684\u5bf9\u6297\u6837\u672c\u653b\u51fb\u65b9\u6cd5\u6765\u63d0\u9ad8\u5bf9\u6297\u6837\u672c\u7684\u53ef\u8fc1\u79fb\u6027\u3002\u5728\u6bcf\u4e00\u6b65\u8ba1\u7b97\u68af\u5ea6\u7684\u8fed\u4ee3\u8fc7\u7a0b\u4e2d\uff0c\u4e0d\u76f4\u63a5\u4f7f\u7528\u5f53\u524d\u68af\u5ea6\uff0c\u800c\u662f\u4f7f\u7528\u5c40\u90e8\u5e73\u5747\u68af\u5ea6\u6765\u7d2f\u79ef\u52a8\u91cf\uff0c\u4ee5\u6b64\u6765\u6291\u5236\u635f\u5931\u51fd\u6570\u66f2\u9762\u5b58\u5728\u7684\u5c40\u90e8\u632f\u8361\u73b0\u8c61\uff0c\u4ece\u800c\u7a33\u5b9a\u66f4\u65b0\u65b9\u5411\uff0c\u9003\u79bb\u5c40\u90e8\u6781\u503c\u70b9\u3002\u5728ImageNet\u6570\u636e\u96c6\u4e0a\u7684\u5927\u91cf\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff1a\u6240\u63d0\u65b9\u6cd5\u4e0e\u73b0\u6709\u57fa\u4e8e\u52a8\u91cf\u7684\u65b9\u6cd5\u76f8\u6bd4\uff0c\u5728\u5355\u4e2a\u6a21\u578b\u653b\u51fb\u5b9e\u9a8c\u4e2d\u7684\u5e73\u5747\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387\u5206\u522b\u63d0\u5347\u4e8638.07%\u548c27.77%\uff0c\u5728\u96c6\u6210\u6a21\u578b\u653b\u51fb\u5b9e\u9a8c\u4e2d\u7684\u5e73\u5747\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387\u5206\u522b\u63d0\u5347\u4e8632.50%\u548c28.63%\u3002" + }, + { + "name": "\u56fe\u50cf\u5185\u5bb9\u7cbe\u7ec6\u5316\u611f\u77e5\u53ca\u5176\u5b89\u5168\u5173\u952e\u6280\u672f\u7814\u7a76", + "authors": [ + "\u738b\u854a1,2", + "\u8346\u4e3d\u68661,2", + "\u90b9\u806a1,2", + "\u5415\u98de\u97041,2", + "\u6731\u5b50\u74871,2" + ], + "affiliations": [ + "1. \u4e2d\u56fd\u79d1\u5b66\u9662\u4fe1\u606f\u5de5\u7a0b\u7814\u7a76\u6240", + "2. \u4e2d\u56fd\u79d1\u5b66\u9662\u5927\u5b66\u7f51\u7edc\u7a7a\u95f4\u5b89\u5168\u5b66\u9662" + ], + "abstract": "\u56fe\u50cf\u5185\u5bb9\u7cbe\u7ec6\u5316\u611f\u77e5\u662f\u8ba1\u7b97\u673a\u89c6\u89c9\u9886\u57df\u5185\u7684\u4e00\u4e2a\u57fa\u7840\u6027\u95ee\u9898,\u65e8\u5728\u5bf9\u56fe\u50cf\u4e2d\u5305\u542b\u7684\u4fe1\u606f\u8fdb\u884c\u7cbe\u7ec6\u5316\u7406\u89e3,\u5177\u6709\u91cd\u8981\u7684\u7814\u7a76\u4ef7\u503c\u548c\u5e7f\u9614\u7684\u5e94\u7528\u573a\u666f\u3002\u6839\u636e\u5173\u6ce8\u8303\u56f4\u7684\u4e0d\u540c,\u56fe\u50cf\u5185\u5bb9\u7cbe\u7ec6\u5316\u611f\u77e5\u4e3b\u8981\u5305\u62ec\u7ec6\u7c92\u5ea6\u8bc6\u522b\u3001\u573a\u666f\u56fe\u751f\u6210\u548c\u56fe\u50cf\u63cf\u8ff0\u7b49\u65b9\u9762\u3002\u672c\u6587\u9996\u5148\u5bf9\u5404\u5173\u952e\u6280\u672f\u7684\u7814\u7a76\u8fdb\u5c55\u548c\u73b0\u72b6\u8fdb\u884c\u7efc\u8ff0;\u7136\u540e\u8ba8\u8bba\u4e86\u76f4\u63a5\u5f71\u54cd\u611f\u77e5\u6a21\u578b\u9884\u6d4b\u7ed3\u679c\u7684\u5b89\u5168\u5a01\u80c1,\u6982\u8ff0\u4e86\u76f8\u5173\u653b\u51fb\u53ca\u9632\u5fa1\u6280\u672f\u7684\u7814\u7a76\u8fdb\u5c55;\u6700\u540e\u5bf9\u8be5\u9886\u57df\u7684\u672a\u6765\u53d1\u5c55\u8d8b\u52bf\u4f5c\u51fa\u5c55\u671b\u3002" + }, + { + "name": "\u878d\u5408\u7f16\u7801\u53ca\u5bf9\u6297\u653b\u51fb\u7684\u5143\u8def\u5f84\u805a\u5408\u56fe\u795e\u7ecf\u7f51\u7edc", + "authors": [ + "\u9648\u5b66\u521a1", + "\u59dc\u5f81\u548c2", + "\u674e\u4f73\u73893" + ], + "affiliations": [ + "1. \u534e\u5317\u7535\u529b\u5927\u5b66\u6570\u7406\u5b66\u9662", + "2. \u667a\u8005\u56db\u6d77(\u5317\u4eac)\u6280\u672f\u6709\u9650\u516c\u53f8", + "3. \u534e\u5317\u7535\u529b\u5927\u5b66\u63a7\u5236\u4e0e\u8ba1\u7b97\u673a\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u5f02\u8d28\u4fe1\u606f\u7f51\u7edc\uff08HIN\uff09\u7531\u4e8e\u5305\u542b\u4e0d\u540c\u7c7b\u578b\u7684\u8282\u70b9\u548c\u8fb9\uff0c \u5728\u5b9e\u9645\u95ee\u9898\u4e2d\u5177\u6709\u5e7f\u6cdb\u7684\u5e94\u7528\u524d\u666f. HIN \u7684\u8868\u793a\u5b66\u4e60\u6a21\u578b\u65e8\u5728\u5bfb\u627e\u4e00\u79cd\u6709\u6548\u7684\u5efa\u6a21\u65b9\u6cd5\uff0c \u5c06 HIN \u4e2d\u7684\u8282\u70b9\u8868\u793a\u4e3a\u4f4e\u7ef4\u5411\u91cf\uff0c \u5e76\u5c3d\u53ef\u80fd\u5730\u4fdd\u7559\u7f51\u7edc\u4e2d\u7684\u5f02\u8d28\u4fe1\u606f. \u7136\u800c\uff0c \u73b0\u6709\u7684\u8868\u793a\u5b66\u4e60\u6a21\u578b\u4ecd\u5b58\u5728\u7740\u5bf9\u5f02\u8d28\u4fe1\u606f\u5229\u7528\u4e0d\u5145\u5206\u7684\u60c5\u51b5. \u4e3a\u89e3\u51b3\u8fd9\u4e00\u95ee\u9898\uff0c \u672c\u6587\u63d0\u51fa\u4e86\u4e00\u79cd\u878d\u5408\u7f16\u7801\u548c\u5bf9\u6297\u653b\u51fb\u7684\u5143\u8def\u5f84\u805a\u5408\u56fe\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\uff08FAMAGNN\uff09\uff0c \u8be5\u6a21\u578b\u7531\u4e09\u4e2a\u6a21\u5757\u6784\u6210\uff0c \u5206\u522b\u662f\u8282\u70b9\u5185\u5bb9\u8f6c\u6362\u3001\u5143\u8def\u5f84\u5185\u805a\u5408\u548c\u5143\u8def\u5f84\u95f4\u805a\u5408. \u8be5\u6a21\u578b\u65e8\u5728\u89e3\u51b3\u73b0\u6709 HIN \u8868\u793a\u5b66\u4e60\u65b9\u6cd5\u63d0\u53d6\u7279\u5f81\u4e0d\u5145\u5206\u7684\u95ee\u9898. \u540c\u65f6\uff0c FAMAGNN\u5f15\u5165\u4e86\u878d\u5408\u7684\u5143\u8def\u5f84\u5b9e\u4f8b\u7f16\u7801\u5668\uff0c \u4ee5\u63d0\u53d6 HIN \u4e2d\u4e30\u5bcc\u7684\u7ed3\u6784\u548c\u8bed\u4e49\u4fe1\u606f. \u6b64\u5916\uff0c \u6a21\u578b\u8fd8\u5f15\u5165\u4e86\u5bf9\u6297\u8bad\u7ec3\uff0c \u5728\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u8fdb\u884c\u5bf9\u6297\u653b\u51fb\uff0c \u4ee5\u63d0\u9ad8\u6a21\u578b\u7684\u9c81\u68d2\u6027. FAMAGNN \u5728\u8282\u70b9\u5206\u7c7b\u548c\u8282\u70b9\u805a\u7c7b\u7b49\u4e0b\u6e38\u4efb\u52a1\u4e2d\u7684\u4f18\u5f02\u8868\u73b0\u8bc1\u660e\u4e86\u5176\u6709\u6548\u6027." + }, + { + "name": "\u9762\u5411\u7f51\u7edc\u5165\u4fb5\u68c0\u6d4b\u7684\u5bf9\u6297\u653b\u51fb\u7cfb\u7edf", + "authors": [ + "\u6f58\u5b87\u6052", + "\u5ed6\u601d\u8d24", + "\u6768\u671d\u4fca", + "\u674e\u5b97\u548c", + "\u4e8e\u5a77\u5a77", + "\u5f20\u745e\u971e" + ], + "affiliations": [ + "\u6842\u6797\u7535\u5b50\u79d1\u6280\u5927\u5b66" + ], + "abstract": "\u8be5\u9879\u76ee\u7814\u7a76\u591a\u79cd\u767d\u76d2\u653b\u51fb\u7b97\u6cd5\u751f\u6210\u767d\u76d2\u5bf9\u6297\u6837\u672c\u7684\u6548\u7387\uff0c\u540c\u65f6\u8fd0\u7528\u751f\u6210\u5bf9\u6297\u7f51\u7edc(GAN)\u6280\u672f\u6765\u751f\u6210\u9ed1\u76d2\u5bf9\u6297\u6837\u672c\uff0c\u5e76\u4e14\u901a\u8fc7\u6784\u5efa\u7f51\u7edc\u5165\u4fb5\u68c0\u6d4b\u6a21\u578b\u5305\u62ec\u8bef\u7528\u68c0\u6d4b\u548c\u5f02\u5e38\u68c0\u6d4b\u6a21\u578b\uff0c\u6765\u6d4b\u8bd5\u8fd9\u4e9b\u751f\u6210\u7684\u5bf9\u6297\u6837\u672c\u5728\u9762\u5bf9\u591a\u79cd\u4e0d\u540c\u5165\u4fb5\u68c0\u6d4b\u6a21\u578b\u65f6\u7684\u653b\u51fb\u7684\u6210\u529f\u7387,\u6700\u7ec8\u6784\u5efa\u4e00\u4e2a\u7f51\u7edc\u5165\u4fb5\u68c0\u6d4b\u7cfb\u7edf\u7684\u5bf9\u6297\u6837\u672c\u751f\u6210\u5668\uff08\u5305\u542b\u767d\u76d2\u548c\u9ed1\u76d2\u5bf9\u6297\u6837\u672c\uff09\u3002" + }, + { + "name": "\u4eba\u8138\u6df1\u5ea6\u4f2a\u9020\u4e3b\u52a8\u9632\u5fa1\u6280\u672f\u7efc\u8ff0", + "authors": [ + "\u77bf\u5de6\u73c9", + "\u6bb7\u742a\u6797", + "\u76db\u7d2b\u7426", + "\u5434\u4fca\u5f66", + "\u5f20\u535a\u6797", + "\u4f59\u5c1a\u620e", + "\u5362\u4f1f" + ], + "affiliations": [], + "abstract": "\u6df1\u5ea6\u751f\u6210\u6a21\u578b\u7684\u98de\u901f\u53d1\u5c55\u63a8\u52a8\u4e86\u4eba\u8138\u6df1\u5ea6\u4f2a\u9020\u6280\u672f\u7684\u8fdb\u6b65\uff0c\u4ee5Deepfake\u4e3a\u4ee3\u8868\u7684\u6df1\u5ea6\u4f2a\u9020\u6a21\u578b\u4e5f\u5f97\u5230\u4e86\u5341\u5206\u5e7f\u6cdb\u7684\u5e94\u7528\u3002\u6df1\u5ea6\u4f2a\u9020\u6280\u672f\u53ef\u4ee5\u5bf9\u4eba\u8138\u56fe\u50cf\u6216\u89c6\u9891\u8fdb\u884c\u6709\u76ee\u7684\u7684\u64cd\u7eb5\uff0c\u4e00\u65b9\u9762\uff0c\u8fd9\u79cd\u6280\u672f\u5e7f\u6cdb\u5e94\u7528\u4e8e\u7535\u5f71\u7279\u6548\u3001\u5a31\u4e50\u573a\u666f\u4e2d\uff0c\u4e30\u5bcc\u4e86\u4eba\u4eec\u7684\u5a31\u4e50\u751f\u6d3b\uff0c\u4fc3\u8fdb\u4e86\u4e92\u8054\u7f51\u591a\u5a92\u4f53\u7684\u4f20\u64ad\uff1b\u53e6\u4e00\u65b9\u9762\uff0c\u6df1\u5ea6\u4f2a\u9020\u4e5f\u5e94\u7528\u4e8e\u4e00\u4e9b\u53ef\u80fd\u9020\u6210\u4e0d\u826f\u5f71\u54cd\u7684\u573a\u666f\uff0c\u7ed9\u516c\u6c11\u7684\u540d\u8a89\u6743\u3001\u8096\u50cf\u6743\u9020\u6210\u4e86\u5371\u5bb3\uff0c\u540c\u65f6\u4e5f\u7ed9\u56fd\u5bb6\u5b89\u5168\u548c\u793e\u4f1a\u7a33\u5b9a\u5e26\u6765\u4e86\u6781\u5927\u7684\u5a01\u80c1\uff0c\u56e0\u6b64\u5bf9\u6df1\u5ea6\u4f2a\u9020\u9632\u5fa1\u6280\u672f\u7684\u7814\u7a76\u65e5\u76ca\u8feb\u5207\u3002\u73b0\u6709\u7684\u9632\u5fa1\u6280\u672f\u4e3b\u8981\u5206\u4e3a\u88ab\u52a8\u68c0\u6d4b\u548c\u4e3b\u52a8\u9632\u5fa1\uff0c\u800c\u88ab\u52a8\u68c0\u6d4b\u7684\u65b9\u5f0f\u65e0\u6cd5\u6d88\u9664\u4f2a\u9020\u4eba\u8138\u5728\u5e7f\u6cdb\u4f20\u64ad\u4e2d\u9020\u6210\u7684\u5f71\u54cd\uff0c\u96be\u4ee5\u505a\u5230\u201c\u4e8b\u524d\u9632\u5fa1\u201d\uff0c\u56e0\u6b64\u4e3b\u52a8\u9632\u5fa1\u7684\u601d\u60f3\u5f97\u5230\u4e86\u7814\u7a76\u4eba\u5458\u7684\u5e7f\u6cdb\u5173\u6ce8\u3002\u7136\u800c\uff0c\u76ee\u524d\u5b66\u672f\u754c\u6709\u5173\u6df1\u5ea6\u4f2a\u9020\u9632\u5fa1\u7684\u7efc\u8ff0\u4e3b\u8981\u5173\u6ce8\u57fa\u4e8e\u68c0\u6d4b\u7684\u88ab\u52a8\u5f0f\u9632\u5fa1\u65b9\u6cd5\uff0c\u51e0\u4e4e\u6ca1\u6709\u4ee5\u6df1\u5ea6\u4f2a\u9020\u4e3b\u52a8\u9632\u5fa1\u6280\u672f\u4e3a\u91cd\u70b9\u7684\u7efc\u8ff0\u3002\u57fa\u4e8e\u6b64\uff0c\u672c\u6587\u5bf9\u5f53\u524d\u5b66\u672f\u754c\u63d0\u51fa\u7684\u4eba\u8138\u6df1\u5ea6\u4f2a\u9020\u4e3b\u52a8\u9632\u5fa1\u6280\u672f\u8fdb\u884c\u68b3\u7406\u3001\u603b\u7ed3\u548c\u8ba8\u8bba\u3002\u9996\u5148\u9610\u8ff0\u4e86\u6df1\u5ea6\u4f2a\u9020\u4e3b\u52a8\u9632\u5fa1\u7684\u63d0\u51fa\u80cc\u666f\u548c\u4e3b\u8981\u601d\u60f3\uff0c\u5e76\u5bf9\u73b0\u6709\u7684\u4eba\u8138\u6df1\u5ea6\u4f2a\u9020\u4e3b\u52a8\u9632\u5fa1\u7b97\u6cd5\u8fdb\u884c\u6c47\u603b\u548c\u5f52\u7c7b\uff0c\u7136\u540e\u5bf9\u5404\u7c7b\u4e3b\u52a8\u9632\u5fa1\u7b97\u6cd5\u7684\u6280\u672f\u539f\u7406\u3001\u6027\u80fd\u3001\u4f18\u7f3a\u70b9\u7b49\u8fdb\u884c\u4e86\u7cfb\u7edf\u6027\u7684\u603b\u7ed3\uff0c\u540c\u65f6\u4ecb\u7ecd\u4e86\u7814\u7a76\u5e38\u7528\u7684\u6570\u636e\u96c6\u548c\u8bc4\u4f30\u65b9\u6cd5\uff0c\u6700\u540e\u5bf9\u6df1\u5ea6\u4f2a\u9020\u4e3b\u52a8\u9632\u5fa1\u6240\u9762\u4e34\u7684\u6280\u672f\u6311\u6218\u8fdb\u884c\u4e86\u5206\u6790\uff0c\u5bf9\u5176\u672a\u6765\u7684\u53d1\u5c55\u65b9\u5411\u5c55\u5f00\u4e86\u601d\u8003\u548c\u8ba8\u8bba\u3002 " + }, + { + "name": "\u57fa\u4e8e\u5c40\u90e8\u6270\u52a8\u7684\u65f6\u95f4\u5e8f\u5217\u9884\u6d4b\u5bf9\u6297\u653b\u51fb", + "authors": [ + "\u5f20\u8000\u51431,2", + "\u539f\u7ee7\u4e1c1,2", + "\u5218\u6d77\u6d0b2", + "\u738b\u5fd7\u6d772", + "\u8d75\u57f9\u7fd42" + ], + "affiliations": [ + "1. \u4ea4\u901a\u5927\u6570\u636e\u4e0e\u4eba\u5de5\u667a\u80fd\u6559\u80b2\u90e8\u91cd\u70b9\u5b9e\u9a8c\u5ba4(\u5317\u4eac\u4ea4\u901a\u5927\u5b66)", + "2. \u5317\u4eac\u4ea4\u901a\u5927\u5b66\u8ba1\u7b97\u673a\u4e0e\u4fe1\u606f\u6280\u672f\u5b66\u9662" + ], + "abstract": "\u65f6\u95f4\u5e8f\u5217\u9884\u6d4b\u6a21\u578b\u5df2\u5e7f\u6cdb\u5e94\u7528\u4e8e\u65e5\u5e38\u751f\u6d3b\u4e2d\u7684\u5404\u4e2a\u884c\u4e1a,\u9488\u5bf9\u8fd9\u4e9b\u9884\u6d4b\u6a21\u578b\u7684\u5bf9\u6297\u653b\u51fb\u5173\u7cfb\u5230\u5404\u884c\u4e1a\u6570\u636e\u7684\u5b89\u5168\u6027.\u76ee\u524d,\u65f6\u95f4\u5e8f\u5217\u7684\u5bf9\u6297\u653b\u51fb\u591a\u5728\u5168\u5c40\u8303\u56f4\u5185\u8fdb\u884c\u5927\u89c4\u6a21\u6270\u52a8,\u5bfc\u81f4\u5bf9\u6297\u6837\u672c\u6613\u88ab\u611f\u77e5.\u540c\u65f6,\u5bf9\u6297\u653b\u51fb\u7684\u6548\u679c\u4f1a\u968f\u7740\u6270\u52a8\u5e45\u5ea6\u7684\u964d\u4f4e\u800c\u660e\u663e\u4e0b\u964d.\u56e0\u6b64,\u5982\u4f55\u5728\u751f\u6210\u4e0d\u6613\u5bdf\u89c9\u7684\u5bf9\u6297\u6837\u672c\u7684\u540c\u65f6\u4fdd\u6301\u8f83\u597d\u7684\u653b\u51fb\u6548\u679c,\u662f\u5f53\u524d\u65f6\u95f4\u5e8f\u5217\u9884\u6d4b\u5bf9\u6297\u653b\u51fb\u9886\u57df\u4e9f\u9700\u89e3\u51b3\u7684\u95ee\u9898\u4e4b\u4e00.\u9996\u5148\u63d0\u51fa\u4e00\u79cd\u57fa\u4e8e\u6ed1\u52a8\u7a97\u53e3\u7684\u5c40\u90e8\u6270\u52a8\u7b56\u7565,\u7f29\u5c0f\u5bf9\u6297\u6837\u672c\u7684\u6270\u52a8\u533a\u95f4;\u5176\u6b21,\u4f7f\u7528\u5dee\u5206\u8fdb\u5316\u7b97\u6cd5\u5bfb\u627e\u6700\u4f18\u653b\u51fb\u70b9\u4f4d,\u5e76\u7ed3\u5408\u5206\u6bb5\u51fd\u6570\u5206\u5272\u6270\u52a8\u533a\u95f4,\u8fdb\u4e00\u6b65\u964d\u4f4e\u6270\u52a8\u8303\u56f4,\u5b8c\u6210\u534a\u767d\u76d2\u653b\u51fb.\u548c\u5df2\u6709\u7684\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\u5728\u591a\u4e2a\u4e0d\u540c\u6df1\u5ea6\u6a21\u578b\u4e0a\u7684\u5bf9\u6bd4\u5b9e\u9a8c\u8868\u660e,\u6240\u63d0\u51fa\u7684\u65b9\u6cd5\u80fd\u591f\u751f\u6210\u4e0d\u6613\u611f\u77e5\u7684\u5bf9\u6297\u6837\u672c,\u5e76\u6709\u6548\u6539\u53d8\u6a21\u578b\u7684\u9884\u6d4b\u8d8b\u52bf,\u5728\u80a1\u7968\u4ea4\u6613\u3001\u7535\u529b\u6d88\u8017\u3001\u592a\u9633\u9ed1\u5b50\u89c2\u6d4b\u548c\u6c14\u6e29\u9884\u6d4b\u8fd94\u4e2a\u5177\u6709\u6311\u6218\u6027\u7684\u4efb\u52a1\u4e2d\u5747\u53d6\u5f97\u4e86\u8f83\u597d\u7684\u653b\u51fb\u6548\u679c." + }, + { + "name": "\u6587\u672c\u5bf9\u6297\u653b\u9632\u6280\u672f\u5728\u7535\u4fe1\u7f51\u7edc\u8bc8\u9a97\u9632\u63a7\u9886\u57df\u7684\u5e94\u7528\u63a2\u6790", + "authors": [ + "\u6c64\u535a\u6587" + ], + "affiliations": [ + "\u4e2d\u56fd\u4eba\u6c11\u8b66\u5bdf\u5927\u5b66(\u5eca\u574a)" + ], + "abstract": "\u968f\u7740\u81ea\u7136\u8bed\u8a00\u5904\u7406\u6a21\u578b\u8fd1\u671f\u5728\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u201c\u51fa\u5708\u201d\uff0c\u6838\u5fc3\u6a21\u578b\u6d89\u53ca\u7684\u5bf9\u6297\u653b\u9632\u6280\u672f\u7684\u53d1\u5c55\uff0c\u9010\u6e10\u6210\u4e3a\u4e00\u628a\u201c\u53cc\u5203\u5251\u201d\uff0c\u7535\u4fe1\u7f51\u7edc\u8bc8\u9a97\u4e0e\u9632\u63a7\u9886\u57df\u53cc\u65b9\u7684\u535a\u5f08\u5bf9\u6297\u6210\u4e3a\u7814\u7a76\u70ed\u70b9\u95ee\u9898\u3002\u901a\u8fc7\u5bf9\u4e0d\u540c\u8bc8\u9a97\u7c7b\u578b\u8fdb\u884c\u5206\u6790\uff0c\u7ed3\u5408\u7535\u4fe1\u7f51\u7edc\u8bc8\u9a97\u5168\u94fe\u6761\u4e0e\u73b0\u6709\u9632\u63a7\u6a21\u578b\u7684\u95ee\u9898\uff0c\u6df1\u5165\u6316\u6398\u53cd\u8bc8\u5e73\u53f0\u6838\u5fc3\u6280\u672f\uff0c\u8bbe\u8ba1\u4e86\u9488\u5bf9\u53cd\u8bc8\u68c0\u6d4b\u6a21\u578b\u7684\u6a21\u62df\u6587\u672c\u5bf9\u6297\u653b\u51fb\uff0c\u63a2\u6790\u6587\u672c\u5bf9\u6297\u653b\u9632\u6280\u672f\u5728\u7535\u4fe1\u7f51\u7edc\u8bc8\u9a97\u9632\u63a7\u9886\u57df\u7684\u5e94\u7528\uff0c\u5e76\u4e14\u8ba8\u8bba\u5176\u9762\u4e34\u7684\u6311\u6218\u4e0e\u524d\u666f\u3002" + }, + { + "name": "\u4e00\u79cd\u968f\u673a\u675f\u641c\u7d22\u6587\u672c\u653b\u51fb\u9ed1\u76d2\u7b97\u6cd5", + "authors": [ + "\u738b\u5c0f\u840c", + "\u5f20\u534e", + "\u4e01\u91d1\u6263", + "\u738b\u7a3c\u6167" + ], + "affiliations": [ + "\u5317\u4eac\u90ae\u7535\u5927\u5b66\u7f51\u7edc\u4e0e\u4ea4\u6362\u6280\u672f\u56fd\u5bb6\u91cd\u70b9\u5b9e\u9a8c\u5ba4" + ], + "abstract": "\u9488\u5bf9\u73b0\u6709\u7684\u5bf9\u6297\u6587\u672c\u751f\u6210\u7b97\u6cd5\u4e2d\u6613\u9677\u5165\u5c40\u90e8\u6700\u4f18\u89e3\u7684\u95ee\u9898\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u5229\u7528\u675f\u641c\u7d22\u548c\u968f\u673a\u5143\u6765\u63d0\u9ad8\u653b\u51fb\u6210\u529f\u7387\u7684R-attack\u7b97\u6cd5\u3002\u9996\u5148\u901a\u8fc7\u5229\u7528\u675f\u641c\u7d22\u6765\u5145\u5206\u5229\u7528\u540c\u4e49\u8bcd\u7a7a\u95f4\u6765\u641c\u7d22\u6700\u4f18\u89e3\uff0c\u4ece\u800c\u589e\u52a0\u751f\u6210\u5bf9\u6297\u6837\u672c\u7684\u591a\u6837\u6027\uff1b\u5e76\u4e14\u5728\u8fed\u4ee3\u641c\u7d22\u8fc7\u7a0b\u4e2d\uff0c\u5f15\u5165\u968f\u673a\u5143\uff0c\u7528\u4e8e\u9632\u6b62\u56e0\u5bfb\u627e\u5bf9\u6297\u6837\u672c\u8fc7\u7a0b\u4e2d\u8fc7\u65e9\u6536\u655b\u800c\u9677\u5165\u5c40\u90e8\u6700\u4f18\u89e3\u7684\u56f0\u5883\u3002\u57283\u4e2a\u6570\u636e\u96c6\u5bf92\u4e2a\u6a21\u578b\u8fdb\u884c\u4e86\u5bf9\u6297\u653b\u51fb\u5b9e\u9a8c\uff0c\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u7528R-attack\u7b97\u6cd5\u80fd\u591f\u6709\u6548\u63d0\u9ad8\u5bf9\u6297\u6837\u672c\u7684\u653b\u51fb\u6210\u529f\u7387\u3002\u4ee5\u653b\u51fbYahoo! Answers\u4e0a\u8bad\u7ec3\u7684LSTM\u6a21\u578b\u4e3a\u4f8b\uff0c\u7528R-attack\u7b97\u6cd5\u653b\u51fb\u6a21\u578b\u7684\u653b\u51fb\u6210\u529f\u7387\u76f8\u6bd4\u57fa\u7ebf\u63d0\u53472.4%\u3002" + }, + { + "name": "\u57fa\u4e8e\u51b3\u7b56\u8fb9\u754c\u654f\u611f\u6027\u548c\u5c0f\u6ce2\u53d8\u6362\u7684\u7535\u78c1\u4fe1\u53f7\u8c03\u5236\u667a\u80fd\u8bc6\u522b\u5bf9\u6297\u6837\u672c\u68c0\u6d4b\u65b9\u6cd5", + "authors": [ + "\u5f90\u4e1c\u4f1f1,2", + "\u848b\u658c1,2", + "\u6731\u6167\u71d51,2", + "\u5ba3\u74261,2", + "\u738b\u5dcd3", + "\u6797\u4e914", + "\u6c88\u4f1f\u56fd3", + "\u6768\u5c0f\u725b1,2,3" + ], + "affiliations": [ + "1. \u6d59\u6c5f\u5de5\u4e1a\u5927\u5b66\u7f51\u7edc\u5b89\u5168\u7814\u7a76\u9662", + "2. \u6d59\u6c5f\u5de5\u4e1a\u5927\u5b66\u4fe1\u606f\u5de5\u7a0b\u5b66\u9662", + "3. \u91cd\u70b9\u7535\u78c1\u7a7a\u95f4\u5b89\u5168\u5168\u56fd\u91cd\u70b9\u5b9e\u9a8c\u5ba4", + "4. \u54c8\u5c14\u6ee8\u5de5\u7a0b\u5927\u5b66\u4fe1\u606f\u4e0e\u901a\u4fe1\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u6df1\u5ea6\u5b66\u4e60\u5728\u56fe\u50cf\u5206\u7c7b\u548c\u5206\u5272\u3001\u7269\u4f53\u68c0\u6d4b\u548c\u8ffd\u8e2a\u3001\u533b\u7597\u3001\u7ffb\u8bd1\u548c\u8bed\u97f3\u8bc6\u522b\u7b49\u4e0e\u4eba\u7c7b\u76f8\u5173\u7684\u4efb\u52a1\u4e2d\u53d6\u5f97\u4e86\u5de8\u5927\u7684\u6210\u529f\u3002\u5b83\u80fd\u591f\u5904\u7406\u5927\u91cf\u590d\u6742\u7684\u6570\u636e\uff0c\u5e76\u81ea\u52a8\u63d0\u53d6\u7279\u5f81\u8fdb\u884c\u9884\u6d4b\uff0c\u56e0\u6b64\u53ef\u4ee5\u66f4\u51c6\u786e\u5730\u9884\u6d4b\u7ed3\u679c\u3002\u968f\u7740\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u7684\u4e0d\u65ad\u53d1\u5c55\uff0c\u4ee5\u53ca\u53ef\u83b7\u5f97\u7684\u6570\u636e\u548c\u8ba1\u7b97\u80fd\u529b\u7684\u63d0\u9ad8\uff0c\u8fd9\u4e9b\u5e94\u7528\u7684\u51c6\u786e\u6027\u4e0d\u65ad\u63d0\u5347\u3002\u6700\u8fd1\uff0c\u6df1\u5ea6\u5b66\u4e60\u4e5f\u5728\u7535\u78c1\u4fe1\u53f7\u9886\u57df\u5f97\u5230\u4e86\u5e7f\u6cdb\u5e94\u7528\uff0c\u4f8b\u5982\u5229\u7528\u795e\u7ecf\u7f51\u7edc\u6839\u636e\u4fe1\u53f7\u7684\u9891\u57df\u548c\u65f6\u57df\u7279\u5f81\u5bf9\u5176\u8fdb\u884c\u5206\u7c7b\u3002\u4f46\u795e\u7ecf\u7f51\u7edc\u5bb9\u6613\u53d7\u5230\u5bf9\u6297\u6837\u672c\u7684\u5e72\u6270\uff0c\u8fd9\u4e9b\u5bf9\u6297\u6837\u672c\u53ef\u4ee5\u8f7b\u6613\u6b3a\u9a97\u795e\u7ecf\u7f51\u7edc\uff0c\u5bfc\u81f4\u5206\u7c7b\u9519\u8bef\u3002\u56e0\u6b64\uff0c\u5bf9\u6297\u6837\u672c\u7684\u751f\u6210\u3001\u68c0\u6d4b\u548c\u9632\u62a4\u7684\u7814\u7a76\u53d8\u5f97\u5c24\u4e3a\u91cd\u8981\uff0c\u8fd9\u5c06\u4fc3\u8fdb\u6df1\u5ea6\u5b66\u4e60\u5728\u7535\u78c1\u4fe1\u53f7\u9886\u57df\u548c\u5176\u4ed6\u9886\u57df\u7684\u53d1\u5c55\u3002\u9488\u5bf9\u73b0\u9636\u6bb5\u5355\u4e00\u7684\u68c0\u6d4b\u65b9\u6cd5\u7684\u6709\u6548\u6027\u4e0d\u9ad8\u7684\u95ee\u9898\uff0c\u63d0\u51fa\u4e86\u57fa\u4e8e\u51b3\u7b56\u8fb9\u754c\u654f\u611f\u6027\u548c\u5c0f\u6ce2\u53d8\u6362\u91cd\u6784\u7684\u5bf9\u6297\u6837\u672c\u68c0\u6d4b\u65b9\u6cd5\u3002\u5229\u7528\u4e86\u5bf9\u6297\u6837\u672c\u4e0e\u6b63\u5e38\u6837\u672c\u5bf9\u6a21\u578b\u51b3\u7b56\u8fb9\u754c\u7684\u654f\u611f\u6027\u5dee\u5f02\u6765\u8fdb\u884c\u68c0\u6d4b\uff0c\u63a5\u7740\u9488\u5bf9\u7b2c\u4e00\u68c0\u6d4b\u9636\u6bb5\u4e2d\u672a\u68c0\u6d4b\u51fa\u7684\u5bf9\u6297\u6837\u672c\uff0c\u672c\u6587\u5229\u7528\u5c0f\u6ce2\u53d8\u6362\u5bf9\u6837\u672c\u8fdb\u884c\u91cd\u6784\uff0c\u5229\u7528\u6837\u672c\u53bb\u566a\u524d\u540e\u5728\u6a21\u578b\u4e2d\u7684\u9884\u6d4b\u503c\u5dee\u5f02\u6765\u8fdb\u884c\u68c0\u6d4b\u3002\u672c\u6587\u5728\u4e24\u79cd\u8c03\u5236\u4fe1\u53f7\u6570\u636e\u96c6\u4e0a\u8fdb\u884c\u4e86\u5b9e\u9a8c\u5206\u6790\uff0c\u5e76\u4e0e\u57fa\u7ebf\u68c0\u6d4b\u65b9\u6cd5\u8fdb\u884c\u5bf9\u6bd4\uff0c\u6b64\u65b9\u6cd5\u66f4\u4f18\u3002\u8fd9\u4e00\u7814\u7a76\u7684\u521b\u65b0\u70b9\u5728\u4e8e\u7efc\u5408\u8003\u8651\u4e86\u6a21\u578b\u51b3\u7b56\u8fb9\u754c\u7684\u654f\u611f\u6027\u548c\u5c0f\u6ce2\u53d8\u6362\u7684\u91cd\u6784\u80fd\u529b\uff0c\u901a\u8fc7\u5de7\u5999\u7684\u7ec4\u5408\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u66f4\u4e3a\u5168\u9762\u3001\u7cbe\u51c6\u7684\u5bf9\u6297\u6837\u672c\u68c0\u6d4b\u65b9\u6cd5\u3002\u8fd9\u4e3a\u6df1\u5ea6\u5b66\u4e60\u5728\u7535\u78c1\u4fe1\u53f7\u9886\u57df\u7684\u7a33\u5065\u5e94\u7528\u63d0\u4f9b\u4e86\u65b0\u7684\u601d\u8def\u548c\u65b9\u6cd5\u3002 " + }, + { + "name": "\u9762\u5411\u667a\u80fd\u65e0\u4eba\u901a\u4fe1\u7cfb\u7edf\u7684\u56e0\u679c\u6027\u5bf9\u6297\u653b\u51fb\u751f\u6210\u7b97\u6cd5", + "authors": [ + "\u79b9\u6811\u65871", + "\u8bb8\u5a011,2", + "\u59da\u5609\u94d61" + ], + "affiliations": [ + "1. \u4e1c\u5357\u5927\u5b66\u79fb\u52a8\u901a\u4fe1\u5168\u56fd\u91cd\u70b9\u5b9e\u9a8c\u5ba4", + "2. \u7f51\u7edc\u901a\u4fe1\u4e0e\u5b89\u5168\u7d2b\u91d1\u5c71\u5b9e\u9a8c\u5ba4" + ], + "abstract": "\u8003\u8651\u5230\u57fa\u4e8e\u68af\u5ea6\u7684\u5bf9\u6297\u653b\u51fb\u751f\u6210\u7b97\u6cd5\u5728\u5b9e\u9645\u901a\u4fe1\u7cfb\u7edf\u90e8\u7f72\u4e2d\u9762\u4e34\u7740\u56e0\u679c\u6027\u95ee\u9898\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u56e0\u679c\u6027\u5bf9\u6297\u653b\u51fb\u751f\u6210\u7b97\u6cd5\u3002\u5229\u7528\u957f\u77ed\u671f\u8bb0\u5fc6\u7f51\u7edc\u7684\u5e8f\u5217\u8f93\u5165\u8f93\u51fa\u7279\u5f81\u4e0e\u65f6\u5e8f\u8bb0\u5fc6\u80fd\u529b\uff0c\u5728\u6ee1\u8db3\u5b9e\u9645\u5e94\u7528\u4e2d\u5b58\u5728\u7684\u56e0\u679c\u6027\u7ea6\u675f\u524d\u63d0\u4e0b\uff0c\u6709\u6548\u63d0\u53d6\u901a\u4fe1\u4fe1\u53f7\u7684\u65f6\u5e8f\u76f8\u5173\u6027\uff0c\u589e\u5f3a\u9488\u5bf9\u65e0\u4eba\u901a\u4fe1\u7cfb\u7edf\u7684\u5bf9\u6297\u653b\u51fb\u6027\u80fd\u3002\u4eff\u771f\u7ed3\u679c\u8868\u660e\uff0c\u6240\u63d0\u7b97\u6cd5\u5728\u540c\u7b49\u6761\u4ef6\u4e0b\u7684\u653b\u51fb\u6027\u80fd\u4f18\u4e8e\u6cdb\u7528\u5bf9\u6297\u6270\u52a8\u7b49\u73b0\u6709\u7684\u56e0\u679c\u6027\u5bf9\u6297\u653b\u51fb\u751f\u6210\u7b97\u6cd5\u3002" + }, + { + "name": "\u57fa\u4e8e\u6f5c\u5728\u6570\u636e\u6316\u6398\u7684\u5c0f\u6837\u672c\u6570\u636e\u5e93\u5bf9\u6297\u653b\u51fb\u9632\u5fa1\u7b97\u6cd5", + "authors": [ + "\u66f9\u537f" + ], + "affiliations": [ + "\u95fd\u5357\u7406\u5de5\u5b66\u9662\u4fe1\u606f\u7ba1\u7406\u5b66\u9662" + ], + "abstract": "\u4e3a\u4e86\u964d\u4f4e\u5c0f\u6837\u672c\u6570\u636e\u5e93\u6b3a\u9a97\u7387\uff0c\u63d0\u5347\u5c0f\u6837\u672c\u6570\u636e\u5e93\u7684\u653b\u51fb\u9632\u5fa1\u6548\u679c\uff0c\u8bbe\u8ba1\u4e86\u4e00\u79cd\u57fa\u4e8e\u6f5c\u5728\u6570\u636e\u6316\u6398\u7684\u5c0f\u6837\u672c\u6570\u636e\u5e93\u5bf9\u6297\u653b\u51fb\u7684\u9632\u5fa1\u7b97\u6cd5(\u6f5c\u5728\u6570\u636e\u6316\u6398\u7684\u9632\u5fa1\u7b97\u6cd5).\u91c7\u7528\u6539\u8fdb\u7684Apriori\u7b97\u6cd5\uff0c\u901a\u8fc7\u9891\u7e41\u5c5e\u6027\u503c\u96c6\u7684\u5de5\u4f5c\u8fc7\u7a0b\u83b7\u53d6\u51c6\u786e\u7684\u5f3a\u5173\u8054\u89c4\u5219\u4f18\u52bf\uff0c\u5e76\u4ece\u5c0f\u6837\u672c\u6570\u636e\u5e93\u4e2d\u6316\u6398\u6f5c\u5728\u6570\u636e\u5bf9\u6297\u653b\u51fb\uff0c\u540c\u65f6\u4f18\u5316\u5019\u9009\u96c6\u5bfb\u627e\u9891\u7e41\u96c6\u7684\u8fc7\u7a0b\uff0c\u7136\u540e\u5229\u7528\u5173\u8054\u5206\u6790\u68c0\u6d4b\u5bf9\u6297\u653b\u51fb\uff0c\u5e76\u901a\u8fc7\u53ef\u4fe1\u5ea6\u8c03\u5ea6\u63a7\u5236\u8bbf\u95ee\u901f\u7387\u6765\u9632\u6b62\u4ea7\u751f\u6076\u610f\u4f1a\u8bdd\uff0c\u5b9e\u73b0\u5c0f\u6837\u672c\u6570\u636e\u5e93\u5bf9\u6297\u653b\u51fb\u9632\u5fa1.\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u6f5c\u5728\u6570\u636e\u6316\u6398\u7684\u9632\u5fa1\u7b97\u6cd5\u53ef\u6709\u6548\u9632\u5fa1\u5c0f\u6837\u672c\u6570\u636e\u5e93\u906d\u53d7\u7684\u591a\u79cd\u7c7b\u578b\u653b\u51fb\uff0c\u964d\u4f4e\u653b\u51fb\u4ea7\u751f\u7684\u6570\u636e\u5e93\u6b3a\u9a97\u7387\uff0c\u4fdd\u969c\u5c0f\u6837\u672c\u6570\u636e\u5e93\u670d\u52a1\u5668\u5229\u7528\u7387\u7684\u7a33\u5b9a\u6027." + }, + { + "name": "\u57fa\u4e8e\u96c5\u53ef\u6bd4\u663e\u8457\u56fe\u7684\u7535\u78c1\u4fe1\u53f7\u5feb\u901f\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5", + "authors": [ + "\u5f20\u5251", + "\u5468\u4fa0", + "\u5f20\u4e00\u7136", + "\u738b\u6893\u806a" + ], + "affiliations": [ + "\u6b66\u6c49\u6570\u5b57\u5de5\u7a0b\u7814\u7a76\u6240" + ], + "abstract": "\u4e3a\u4e86\u751f\u6210\u9ad8\u8d28\u91cf\u7684\u7535\u78c1\u4fe1\u53f7\u5bf9\u6297\u6837\u672c\uff0c\u63d0\u51fa\u4e86\u5feb\u901f\u96c5\u53ef\u6bd4\u663e\u8457\u56fe\u653b\u51fb\uff08FJSMA\uff09\u65b9\u6cd5\u3002FJSMA\u901a\u8fc7\u8ba1\u7b97\u653b\u51fb\u76ee\u6807\u7c7b\u522b\u7684\u96c5\u53ef\u6bd4\u77e9\u9635\uff0c\u5e76\u6839\u636e\u8be5\u77e9\u9635\u751f\u6210\u7279\u5f81\u663e\u8457\u56fe\uff0c\u4e4b\u540e\u8fed\u4ee3\u9009\u53d6\u663e\u8457\u6027\u6700\u5f3a\u7684\u7279\u5f81\u70b9\u53ca\u5176\u90bb\u57df\u5185\u8fde\u7eed\u7279\u5f81\u70b9\u6dfb\u52a0\u6270\u52a8\uff0c\u540c\u65f6\u5f15\u5165\u5355\u70b9\u6270\u52a8\u9650\u5236\uff0c\u6700\u540e\u751f\u6210\u5bf9\u6297\u6837\u672c\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u4e0e\u96c5\u53ef\u6bd4\u663e\u8457\u56fe\u653b\u51fb\u65b9\u6cd5\u76f8\u6bd4\uff0cFJSMA\u5728\u4fdd\u6301\u4e0e\u4e4b\u76f8\u540c\u7684\u9ad8\u653b\u51fb\u6210\u529f\u7387\u7684\u540c\u65f6\uff0c\u751f\u6210\u901f\u5ea6\u63d0\u5347\u4e86\u7ea610\u500d\uff0c\u76f8\u4f3c\u5ea6\u63d0\u5347\u4e86\u8d85\u8fc711%\uff1b\u4e0e\u5176\u4ed6\u57fa\u4e8e\u68af\u5ea6\u7684\u65b9\u6cd5\u76f8\u6bd4\uff0c\u653b\u51fb\u6210\u529f\u7387\u63d0\u5347\u4e86\u8d85\u8fc720%\uff0c\u76f8\u4f3c\u5ea6\u63d0\u5347\u4e8620%\uff5e30%\u3002" + }, + { + "name": "\u57fa\u4e8e\u52a8\u91cf\u8fed\u4ee3\u5feb\u901f\u68af\u5ea6\u7b26\u53f7\u7684SAR-ATR\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5", + "authors": [ + "\u4e07\u70dc\u7533", + "\u5218\u4f1f", + "\u725b\u671d\u9633", + "\u5362\u4e07\u6770" + ], + "affiliations": [ + "\u4e2d\u56fd\u4eba\u6c11\u89e3\u653e\u519b\u6218\u7565\u652f\u63f4\u90e8\u961f\u4fe1\u606f\u5de5\u7a0b\u5927\u5b66\u6570\u636e\u4e0e\u76ee\u6807\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u5408\u6210\u5b54\u5f84\u96f7\u8fbe\u81ea\u52a8\u76ee\u6807\u8bc6\u522b(SAR-ATR)\u9886\u57df\u7f3a\u4e4f\u6709\u6548\u7684\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5\uff0c\u4e3a\u6b64\uff0c\u8be5\u6587\u7ed3\u5408\u52a8\u91cf\u8fed\u4ee3\u5feb\u901f\u68af\u5ea6\u7b26\u53f7(MI-FGSM)\u601d\u60f3\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u8fc1\u79fb\u7684\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5\u3002\u9996\u5148\u7ed3\u5408SAR\u56fe\u50cf\u7279\u6027\u8fdb\u884c\u968f\u673a\u6591\u70b9\u566a\u58f0\u53d8\u6362\uff0c\u7f13\u89e3\u6a21\u578b\u5bf9\u6591\u70b9\u566a\u58f0\u7684\u8fc7\u62df\u5408\uff0c\u63d0\u9ad8\u7b97\u6cd5\u7684\u6cdb\u5316\u6027\u80fd\uff1b\u7136\u540e\u8bbe\u8ba1\u4e86\u80fd\u591f\u5feb\u901f\u5bfb\u627e\u6700\u4f18\u68af\u5ea6\u4e0b\u964d\u65b9\u5411\u7684ABN\u5bfb\u4f18\u5668\uff0c\u901a\u8fc7\u6a21\u578b\u68af\u5ea6\u5feb\u901f\u6536\u655b\u63d0\u5347\u7b97\u6cd5\u653b\u51fb\u6709\u6548\u6027\uff1b\u6700\u540e\u5f15\u5165\u62df\u53cc\u66f2\u52a8\u91cf\u7b97\u5b50\u83b7\u5f97\u7a33\u5b9a\u7684\u6a21\u578b\u68af\u5ea6\u4e0b\u964d\u65b9\u5411\uff0c\u4f7f\u68af\u5ea6\u5728\u5feb\u901f\u6536\u655b\u8fc7\u7a0b\u4e2d\u907f\u514d\u9677\u5165\u5c40\u90e8\u6700\u4f18\uff0c\u8fdb\u4e00\u6b65\u589e\u5f3a\u5bf9\u6297\u6837\u672c\u7684\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387\u3002\u901a\u8fc7\u4eff\u771f\u5b9e\u9a8c\u8868\u660e\uff0c\u4e0e\u73b0\u6709\u7684\u5bf9\u6297\u653b\u51fb\u7b97\u6cd5\u76f8\u6bd4\uff0c\u8be5\u6587\u7b97\u6cd5\u5728MSTAR\u548cFUSAR-Ship\u6570\u636e\u96c6\u4e0a\u5bf9\u4e3b\u6d41\u7684SAR-ATR\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u7684\u96c6\u6210\u6a21\u578b\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387\u5206\u522b\u63d0\u9ad8\u4e863%\uff5e55%\u548c6%\uff5e57.5%\uff0c\u800c\u4e14\u751f\u6210\u7684\u5bf9\u6297\u6837\u672c\u5177\u6709\u9ad8\u5ea6\u7684\u9690\u853d\u6027\u3002" + }, + { + "name": "\u9762\u5411\u56fe\u50cf\u5206\u6790\u9886\u57df\u7684\u9ed1\u76d2\u5bf9\u6297\u653b\u51fb\u6280\u672f\u7efc\u8ff0", + "authors": [ + "\u6b66\u9633", + "\u5218\u9756" + ], + "affiliations": [ + "\u5185\u8499\u53e4\u5927\u5b66\u8ba1\u7b97\u673a\u5b66\u9662" + ], + "abstract": "\u56fe\u50cf\u9886\u57df\u4e0b\u7684\u9ed1\u76d2\u653b\u51fb\uff08Black-box Attack\uff09\u5df2\u6210\u4e3a\u5f53\u524d\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u5bf9\u6297\u653b\u51fb\u9886\u57df\u7684\u70ed\u70b9\u7814\u7a76\u65b9\u5411\u3002\u9ed1\u76d2\u653b\u51fb\u7684\u7279\u70b9\u5728\u4e8e\u4ec5\u5229\u7528\u6a21\u578b\u8f93\u5165\u4e0e\u8f93\u51fa\u7684\u6620\u5c04\u5173\u7cfb\uff0c\u800c\u65e0\u9700\u6a21\u578b\u5185\u90e8\u53c2\u6570\u4fe1\u606f\u53ca\u68af\u5ea6\u4fe1\u606f\uff0c\u901a\u8fc7\u5411\u56fe\u50cf\u6570\u636e\u52a0\u5165\u4eba\u7c7b\u96be\u4ee5\u5bdf\u89c9\u7684\u5fae\u5c0f\u6270\u52a8\uff0c\u8fdb\u800c\u9020\u6210\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\uff08Deep Neural Network\uff0c DNN\uff09\u63a8\u7406\u4e0e\u8bc6\u522b\u5931\u51c6\uff0c\u5bfc\u81f4\u56fe\u50cf\u5206\u6790\u4efb\u52a1\u7684\u51c6\u786e\u7387\u4e0b\u964d\uff0c\u56e0\u6b64\u7531\u9ed1\u76d2\u653b\u51fb\u5f15\u8d77\u7684\u9c81\u68d2\u6027\u95ee\u9898\u6210\u4e3a\u5f53\u524dDNN\u6a21\u578b\u7814\u7a76\u7684\u5173\u952e\u95ee\u9898\u3002\u4e3a\u63d0\u9ad8\u9ed1\u76d2\u653b\u51fb\u5728\u56fe\u50cf\u5206\u6790\u4efb\u52a1\u4e0b\u7684\u653b\u51fb\u6210\u6548\uff0c\u73b0\u6709\u76f8\u5173\u7814\u7a76\u4ee5\u4f4e\u67e5\u8be2\u6b21\u6570\u3001\u4f4e\u6270\u52a8\u5e45\u5ea6\u3001\u9ad8\u653b\u51fb\u6210\u529f\u7387\u4f5c\u4e3a\u4f18\u5316\u76ee\u6807\uff0c\u9488\u5bf9\u4e0d\u540c\u56fe\u50cf\u5206\u6790\u4efb\u52a1\u91c7\u7528\u4e0d\u540c\u7684\u653b\u51fb\u6a21\u5f0f\u4e0e\u8bc4\u4f30\u65b9\u5f0f\u3002\u672c\u6587\u4ee5\u4e3b\u6d41\u7684\u56fe\u50cf\u5206\u6790\u4efb\u52a1\u4e3a\u51fa\u53d1\u70b9\uff0c\u9610\u8ff0\u56fe\u50cf\u5206\u7c7b\u3001\u76ee\u6807\u68c0\u6d4b\u4e0e\u56fe\u50cf\u5206\u5272\u4e09\u7c7b\u4efb\u52a1\u4e2d\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5\u7684\u6838\u5fc3\u601d\u60f3\u548c\u96be\u70b9\uff0c\u603b\u7ed3\u9ed1\u76d2\u5bf9\u6297\u653b\u51fb\u9886\u57df\u4e2d\u7684\u5173\u952e\u6982\u5ff5\u4e0e\u8bc4\u4f30\u6307\u6807\uff0c\u5206\u6790\u4e0d\u540c\u56fe\u50cf\u5206\u6790\u4efb\u52a1\u4e2d\u9ed1\u76d2\u5bf9\u6297\u653b\u51fb\u7684\u5b9e\u73b0\u7b56\u7565\u4e0e\u7814\u7a76\u76ee\u6807\u3002\u9610\u660e\u5404\u4e2a\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5\u95f4\u7684\u5173\u7cfb\u4e0e\u4f18\u52bf\uff0c\u4ece\u653b\u51fb\u6210\u529f\u7387\u3001\u67e5\u8be2\u6b21\u6570\uff0c\u4ee5\u53ca\u76f8\u4f3c\u6027\u5ea6\u91cf\u7b49\u591a\u4e2a\u65b9\u9762\u5bf9\u4e0d\u540c\u7684\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5\u8fdb\u884c\u6027\u80fd\u6bd4\u8f83\uff0c\u4ee5\u63d0\u51fa\u76ee\u524d\u56fe\u50cf\u5206\u6790\u9886\u57df\u4e2d\u9ed1\u76d2\u5bf9\u6297\u653b\u51fb\u4ecd\u5b58\u5728\u7684\u4e3b\u8981\u6311\u6218\u4e0e\u672a\u6765\u7814\u7a76\u65b9\u5411\u3002" + }, + { + "name": "\u7164\u77ff\u4e95\u4e0b\u94bb\u8fdb\u901f\u5ea6\u5f71\u54cd\u56e0\u7d20\u53ca\u5176\u667a\u80fd\u9884\u6d4b\u65b9\u6cd5\u7814\u7a76", + "authors": [ + "\u6234\u5251\u535a1", + "\u738b\u5fe0\u5bbe1", + "\u5f20\u74301", + "\u53f8\u57921", + "\u9b4f\u4e1c1", + "\u5468\u6587\u535a2", + "\u987e\u8fdb\u60521", + "\u90b9\u7b71\u745c1", + "\u5b8b\u96e8\u96e82" + ], + "affiliations": [ + "1. \u4e2d\u56fd\u77ff\u4e1a\u5927\u5b66\u673a\u68b0\u5de5\u7a0b\u5b66\u9662", + "2. \u56db\u5ddd\u822a\u5929\u7cfb\u7edf\u5de5\u7a0b\u7814\u7a76\u6240" + ], + "abstract": "\u5728\u7164\u77ff\u4e95\u4e0b\u94bb\u63a2\u9886\u57df\uff0c\u94bb\u8fdb\u901f\u5ea6(DR)\u662f\u8bc4\u4f30\u94bb\u63a2\u4f5c\u4e1a\u6700\u6709\u6548\u6307\u6807\u4e4b\u4e00\uff0c\u94bb\u901f\u9884\u6d4b\u662f\u5b9e\u73b0\u7164\u77ff\u94bb\u8fdb\u667a\u80fd\u5316\u7684\u524d\u63d0\u6761\u4ef6\uff0c\u5bf9\u4e8e\u4f18\u5316\u94bb\u673a\u94bb\u8fdb\u53c2\u6570\u3001\u964d\u4f4e\u4f5c\u4e1a\u6210\u672c\u3001\u5b9e\u73b0\u5b89\u5168\u9ad8\u6548\u94bb\u63a2\u5177\u6709\u91cd\u8981\u610f\u4e49\u3002\u4e3a\u6b64\uff0c\u63d0\u51fa\u7164\u77ff\u4e95\u4e0b\u94bb\u8fdb\u901f\u5ea6\u5f71\u54cd\u56e0\u7d20\u53ca\u5176\u667a\u80fd\u9884\u6d4b\u65b9\u6cd5\u7814\u7a76\uff0c\u63a2\u7d22\u57fa\u4e8e\u94bb\u538b\u3001\u8f6c\u901f\u3001\u626d\u77e9\u4ee5\u53ca\u94bb\u8fdb\u6df1\u5ea6\u7b49\u5c11\u91cf\u94bb\u673a\u53c2\u6570\u91c7\u7528\u673a\u5668\u5b66\u4e60\u7b97\u6cd5\u5b9e\u73b0\u94bb\u8fdb\u901f\u5ea6\u7cbe\u51c6\u9884\u6d4b\u3002\u9996\u5148\u901a\u8fc7\u5b9e\u9a8c\u5ba4\u5fae\u94bb\u8bd5\u9a8c\uff0c\u6df1\u5165\u5206\u6790\u7164\u5ca9\u529b\u5b66\u6027\u80fd\u3001\u94bb\u538b\u3001\u8f6c\u901f\u548c\u94bb\u8fdb\u6df1\u5ea6\u5bf9\u626d\u77e9\u3001\u94bb\u8fdb\u901f\u5ea6\u5f71\u54cd\u89c4\u5f8b\u3002\u7814\u7a76\u7ed3\u679c\u663e\u793a\uff0c\u5728\u7164\u77ff\u4e95\u4e0b\u94bb\u8fdb\u8fc7\u7a0b\u4e2d\uff0c\u968f\u7740\u94bb\u8fdb\u538b\u529b\u589e\u5927\uff0c\u94bb\u8fdb\u901f\u5ea6\u5448\u9010\u6e10\u5347\u9ad8\u8d8b\u52bf\uff0c\u5728\u8f83\u9ad8\u7684\u8f6c\u901f\u6761\u4ef6\u4e0b\u94bb\u8fdb\u538b\u529b\u5bf9\u94bb\u8fdb\u901f\u5ea6\u5f71\u54cd\u66f4\u52a0\u660e\u663e\uff0c\u8f6c\u901f\u589e\u52a0\u6709\u5229\u4e8e\u63d0\u9ad8\u94bb\u8fdb\u901f\u5ea6\uff0c\u4f46\u8f6c\u901f\u5bf9\u786c\u5ea6\u8f83\u4f4e\u7684\u7164\u5c42\u94bb\u8fdb\u901f\u5ea6\u5f71\u54cd\u66f4\u4e3a\u663e\u8457\uff1b\u7136\u540e\uff0c\u6839\u636e\u7164\u77ff\u4e95\u4e0b\u9632\u51b2\u94bb\u5b54\u73b0\u573a\u6570\u636e\uff0c\u91c7\u7528K-\u8fd1\u90bb(KNN)\u3001\u652f\u6301\u5411\u91cf\u56de\u5f52(SVR)\u548c\u968f\u673a\u68ee\u6797\u56de\u5f52(RFR)\u4e09\u79cd\u4e0d\u540c\u7684\u673a\u5668\u5b66\u4e60\u7b97\u6cd5\u5efa\u7acb\u94bb\u8fdb\u901f\u5ea6\u9884\u6d4b\u6a21\u578b\uff0c\u5e76\u7ed3\u5408\u7c92\u5b50\u7fa4\u7b97\u6cd5(PSO)\u5bf9\u4e09\u79cd\u6a21\u578b\u8d85\u53c2\u6570\u8fdb\u884c\u4f18\u5316\uff0c\u6700\u540e\u5bf9\u6bd4\u5206\u6790PSO-KNN\u3001PSO-SVR\u548cPSO-RFR\u4e09\u79cd\u94bb\u8fdb\u901f\u5ea6\u9884\u6d4b\u6a21\u578b\u9884\u6d4b\u7ed3\u679c\u3002\u7814\u7a76\u7ed3\u679c\u8868\u660e\uff0cPSO-RFR\u6a21\u578b\u51c6\u786e\u6027\u6700\u597d\uff0c\u51b3\u5b9a\u7cfb\u6570R2\u9ad8\u8fbe0.963\uff0c\u5747\u65b9\u8bef\u5deeMSE\u4ec5\u670929.742\uff0c\u800cPSO-SVR\u6a21\u578b\u9c81\u68d2\u6027\u6700\u597d\uff0c\u5728\u5bf9\u6297\u653b\u51fb\u540e\u8bc4\u4ef7\u6307\u6807\u53d8\u5316\u7387\u6700\u5c0f\u3002\u672c\u6587\u7814\u7a76\u6709\u52a9\u4e8e\u5b9e\u73b0\u7164\u77ff\u4e95\u4e0b\u94bb\u8fdb\u901f\u5ea6\u7684\u7cbe\u51c6\u9884\u6d4b\uff0c\u4e3a\u7164\u77ff\u4e95\u4e0b\u667a\u80fd\u94bb\u8fdb\u53c2\u6570\u4f18\u5316\u63d0\u4f9b\u7406\u8bba\u652f\u6491\u3002 " + }, + { + "name": "\u9488\u5bf9\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u7269\u7406\u5bf9\u6297\u653b\u51fb\u7efc\u8ff0", + "authors": [ + "\u8521\u4f1f", + "\u72c4\u661f\u96e8", + "\u848b\u6615\u660a", + "\u738b\u946b", + "\u9ad8\u851a\u6d01" + ], + "affiliations": [ + "\u706b\u7bad\u519b\u5de5\u7a0b\u5927\u5b66\u5bfc\u5f39\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5bb9\u6613\u53d7\u5230\u5bf9\u6297\u6837\u672c\u7684\u5f71\u54cd\uff0c\u5728\u56fe\u50cf\u4e0a\u6dfb\u52a0\u8089\u773c\u4e0d\u53ef\u89c1\u7684\u5fae\u5c0f\u6270\u52a8\u5c31\u53ef\u4ee5\u4f7f\u8bad\u7ec3\u6709\u7d20\u7684\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5931\u7075\u3002\u6700\u8fd1\u7684\u7814\u7a76\u8868\u660e\u8fd9\u79cd\u6270\u52a8\u4e5f\u5b58\u5728\u4e8e\u73b0\u5b9e\u4e16\u754c\u4e2d\u3002\u805a\u7126\u4e8e\u6df1\u5ea6\u5b66\u4e60\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u7269\u7406\u5bf9\u6297\u653b\u51fb\uff0c\u660e\u786e\u4e86\u7269\u7406\u5bf9\u6297\u653b\u51fb\u7684\u6982\u5ff5\uff0c\u5e76\u4ecb\u7ecd\u4e86\u76ee\u6807\u68c0\u6d4b\u7269\u7406\u5bf9\u6297\u653b\u51fb\u7684\u4e00\u822c\u6d41\u7a0b\uff0c\u4f9d\u636e\u653b\u51fb\u4efb\u52a1\u7684\u4e0d\u540c\u4ece\u8f66\u8f86\u68c0\u6d4b\u548c\u884c\u4eba\u68c0\u6d4b\u7efc\u8ff0\u4e86\u8fd1\u5e74\u6765\u4e00\u7cfb\u5217\u9488\u5bf9\u76ee\u6807\u68c0\u6d4b\u7f51\u7edc\u7684\u7269\u7406\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\uff0c\u4ee5\u53ca\u7b80\u5355\u4ecb\u7ecd\u4e86\u5176\u4ed6\u9488\u5bf9\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u653b\u51fb\u3001\u5176\u4ed6\u653b\u51fb\u4efb\u52a1\u548c\u5176\u4ed6\u653b\u51fb\u65b9\u5f0f\u3002\u6700\u540e\uff0c\u8ba8\u8bba\u4e86\u7269\u7406\u5bf9\u6297\u653b\u51fb\u5f53\u524d\u9762\u4e34\u7684\u6311\u6218\uff0c\u5f15\u51fa\u5bf9\u6297\u8bad\u7ec3\u7684\u5c40\u9650\u6027\u5e76\u5c55\u671b\u672a\u6765\u53ef\u80fd\u7684\u53d1\u5c55\u65b9\u5411\u548c\u5e94\u7528\u524d\u666f\u3002" + }, + { + "name": "\u9488\u5bf9\u81ea\u52a8\u9a7e\u9a76\u667a\u80fd\u6a21\u578b\u7684\u653b\u51fb\u4e0e\u9632\u5fa1", + "authors": [ + "\u9a6c\u66681,2", + "\u6c88\u8d851,2", + "\u853a\u741b\u76931,2", + "\u674e\u524d1,2", + "\u738b\u9a9e3", + "\u674e\u74264", + "\u7ba1\u6653\u5b8f1,2" + ], + "affiliations": [ + "1. \u897f\u5b89\u4ea4\u901a\u5927\u5b66\u7535\u5b50\u4e0e\u4fe1\u606f\u5b66\u90e8\u7f51\u7edc\u7a7a\u95f4\u5b89\u5168\u5b66\u9662", + "2. \u667a\u80fd\u7f51\u7edc\u4e0e\u7f51\u7edc\u5b89\u5168\u6559\u80b2\u90e8\u91cd\u70b9\u5b9e\u9a8c\u5ba4(\u897f\u5b89\u4ea4\u901a\u5927\u5b66)", + "3. \u6b66\u6c49\u5927\u5b66\u56fd\u5bb6\u7f51\u7edc\u5b89\u5168\u5b66\u9662", + "4. \u6e05\u534e\u5927\u5b66\u7f51\u7edc\u79d1\u5b66\u4e0e\u7f51\u7edc\u7a7a\u95f4\u7814\u7a76\u9662" + ], + "abstract": "\u8fd1\u5e74\u6765\uff0c\u4ee5\u6df1\u5ea6\u5b66\u4e60\u7b97\u6cd5\u4e3a\u4ee3\u8868\u7684\u4eba\u5de5\u667a\u80fd\u6280\u672f\u4e3a\u4eba\u7c7b\u751f\u4ea7\u751f\u6d3b\u7684\u65b9\u65b9\u9762\u9762\u5e26\u6765\u4e86\u5de8\u5927\u7684\u9769\u65b0\uff0c\u5c24\u5176\u662f\u5728\u81ea\u52a8\u9a7e\u9a76\u9886\u57df\uff0c\u90e8\u7f72\u7740\u81ea\u52a8\u9a7e\u9a76\u7cfb\u7edf\u7684\u667a\u80fd\u6c7d\u8f66\u5df2\u7ecf\u8d70\u8fdb\u4eba\u4eec\u7684\u751f\u6d3b\uff0c\u6210\u4e3a\u4e86\u91cd\u8981\u7684\u751f\u4ea7\u529b\u5de5\u5177\u3002\u7136\u800c\uff0c\u81ea\u52a8\u9a7e\u9a76\u7cfb\u7edf\u4e2d\u7684\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u9762\u4e34\u7740\u6f5c\u5728\u7684\u5b89\u5168\u9690\u60a3\u548c\u98ce\u9669\uff0c\u8fd9\u7ed9\u4eba\u6c11\u7fa4\u4f17\u751f\u547d\u8d22\u4ea7\u5b89\u5168\u5e26\u6765\u4e86\u4e25\u91cd\u5a01\u80c1\u3002\u672c\u6587\u901a\u8fc7\u56de\u987e\u81ea\u52a8\u9a7e\u9a76\u667a\u80fd\u6a21\u578b\u653b\u51fb\u548c\u9632\u5fa1\u7684\u76f8\u5173\u7814\u7a76\u5de5\u4f5c\uff0c\u63ed\u793a\u81ea\u52a8\u9a7e\u9a76\u7cfb\u7edf\u5728\u7269\u7406\u4e16\u754c\u4e0b\u9762\u4e34\u7684\u5b89\u5168\u98ce\u9669\u5e76\u5f52\u7eb3\u603b\u7ed3\u4e86\u76f8\u5e94\u7684\u9632\u5fa1\u5bf9\u7b56\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u672c\u6587\u9996\u5148\u4ecb\u7ecd\u4e86\u5305\u542b\u653b\u51fb\u9762\u3001\u653b\u51fb\u80fd\u529b\u548c\u653b\u51fb\u76ee\u6807\u7684\u81ea\u52a8\u9a7e\u9a76\u7cfb\u7edf\u5b89\u5168\u98ce\u9669\u6a21\u578b\u3002\u5176\u6b21\uff0c\u9762\u5411\u81ea\u52a8\u9a7e\u9a76\u7cfb\u7edf\u7684\u4e09\u4e2a\u5173\u952e\u529f\u80fd\u5c42\u2014\u2014\u4f20\u611f\u5668\u5c42\u3001\u611f\u77e5\u5c42\u548c\u51b3\u7b56\u5c42\uff0c\u672c\u6587\u4f9d\u636e\u53d7\u653b\u51fb\u7684\u667a\u80fd\u6a21\u578b\u548c\u653b\u51fb\u624b\u6bb5\u5f52\u7eb3\u3001\u5206\u6790\u4e86\u5bf9\u5e94\u7684\u653b\u51fb\u65b9\u6cd5\u4ee5\u53ca\u9632\u5fa1\u5bf9\u7b56\uff0c\u5e76\u63a2\u8ba8\u4e86\u73b0\u6709\u65b9\u6cd5\u7684\u5c40\u9650\u6027\u3002\u6700\u540e\uff0c\u672c\u6587\u8ba8\u8bba\u548c\u5c55\u671b\u4e86\u81ea\u52a8\u9a7e\u9a76\u667a\u80fd\u6a21\u578b\u653b\u51fb\u4e0e\u9632\u5fa1\u6280\u672f\u9762\u4e34\u7684\u96be\u9898\u4e0e\u6311\u6218\uff0c\u5e76\u6307\u51fa\u4e86\u672a\u6765\u6f5c\u5728\u7684\u7814\u7a76\u65b9\u5411\u548c\u53d1\u5c55\u8d8b\u52bf." + }, + { + "name": "\u9690\u79c1\u4fdd\u62a4\u7684\u56fe\u50cf\u66ff\u4ee3\u6570\u636e\u751f\u6210\u65b9\u6cd5", + "authors": [ + "\u674e\u5a49\u83b91,2", + "\u5218\u5b66\u82731,2", + "\u6768\u535a1,2" + ], + "affiliations": [ + "1. \u5409\u6797\u5927\u5b66\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u5b66\u9662", + "2. \u5409\u6797\u5927\u5b66\u7b26\u53f7\u8ba1\u7b97\u4e0e\u77e5\u8bc6\u5de5\u7a0b\u6559\u80b2\u90e8\u91cd\u70b9\u5b9e\u9a8c\u5ba4" + ], + "abstract": "\u9488\u5bf9\u73b0\u6709\u56fe\u50cf\u6570\u636e\u96c6\u5b58\u5728\u7684\u9690\u79c1\u4fdd\u62a4\u9700\u6c42\uff0c\u63d0\u51fa\u4e00\u79cd\u56fe\u50cf\u6570\u636e\u96c6\u9690\u79c1\u4fdd\u62a4\u573a\u666f\u53ca\u8be5\u573a\u666f\u4e0b\u9690\u79c1\u4fdd\u62a4\u7684\u56fe\u50cf\u66ff\u4ee3\u6570\u636e\u751f\u6210\u65b9\u6cd5\u3002\u8be5\u573a\u666f\u5229\u7528\u7ecf\u9690\u79c1\u4fdd\u62a4\u65b9\u6cd5\u5904\u7406\u540e\u7684\u66ff\u4ee3\u56fe\u50cf\u6570\u636e\u96c6\u53d6\u4ee3\u539f\u59cb\u56fe\u50cf\u6570\u636e\u96c6\uff0c\u5176\u4e2d\u66ff\u4ee3\u56fe\u50cf\u4e0e\u539f\u59cb\u56fe\u50cf\u4e00\u4e00\u5bf9\u5e94\uff0c\u4eba\u7c7b\u65e0\u6cd5\u8bc6\u522b\u66ff\u4ee3\u56fe\u50cf\u6240\u5c5e\u7c7b\u522b\uff0c\u66ff\u4ee3\u56fe\u50cf\u53ef\u8bad\u7ec3\u73b0\u6709\u7684\u6df1\u5ea6\u5b66\u4e60\u56fe\u50cf\u5206\u7c7b\u7b97\u6cd5\uff0c\u4e14\u5177\u6709\u8f83\u597d\u7684\u5206\u7c7b\u6548\u679c\u3002\u540c\u65f6\u9488\u5bf9\u4e0a\u8ff0\u573a\u666f\uff0c\u6539\u8fdb\u4e86\u57fa\u4e8e\u6295\u5f71\u68af\u5ea6\u4e0b\u964d(PGD:Project Gradient Descent)\u653b\u51fb\u7684\u6570\u636e\u9690\u79c1\u4fdd\u62a4\u65b9\u6cd5\uff0c\u5c06\u539f\u59cbPGD\u653b\u51fb\u76ee\u6807\u7531\u6807\u7b7e\u6539\u4e3a\u56fe\u50cf\uff0c\u5373\u56fe\u50cf\u5bf9\u56fe\u50cf\u7684\u653b\u51fb\uff0c\u5e76\u4f7f\u7528\u7ecf\u8fc7\u5bf9\u6297\u8bad\u7ec3\u7684\u9c81\u68d2\u6a21\u578b\u8fdb\u884c\u56fe\u50cf\u5bf9\u56fe\u50cf\u653b\u51fb\u4f5c\u4e3a\u66ff\u4ee3\u6570\u636e\u7684\u751f\u6210\u65b9\u6cd5\u3002\u5728\u6807\u51c6\u6d4b\u8bd5\u96c6\u4e0a\uff0c\u66ff\u4ee3\u540e\u7684CIFAR(Canadian Institute For Advanced Research 10)\u6570\u636e\u96c6\u548cCINIC\u6570\u636e\u96c6\u5728\u56fe\u50cf\u5206\u7c7b\u4efb\u52a1\u4e0a\u5206\u522b\u53d6\u5f97\u4e8687.15%\u548c74.04%\u7684\u6d4b\u8bd5\u6b63\u786e\u7387\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u8be5\u65b9\u6cd5\u80fd\u5728\u4fdd\u8bc1\u66ff\u4ee3\u6570\u636e\u96c6\u5bf9\u4eba\u7c7b\u9690\u79c1\u6027\u7684\u524d\u63d0\u4e0b\uff0c\u751f\u6210\u539f\u59cb\u6570\u636e\u96c6\u7684\u66ff\u4ee3\u6570\u636e\u96c6\uff0c\u5e76\u4fdd\u8bc1\u73b0\u6709\u65b9\u6cd5\u5728\u8be5\u6570\u636e\u96c6\u4e0a\u7684\u5206\u7c7b\u6027\u80fd\u3002" + }, + { + "name": "\u7ed3\u5408\u81ea\u9002\u5e94\u6b65\u957f\u7b56\u7565\u548c\u6570\u636e\u589e\u5f3a\u673a\u5236\u63d0\u5347\u5bf9\u6297\u653b\u51fb\u8fc1\u79fb\u6027", + "authors": [ + "\u9c8d\u857e1", + "\u9676\u851a2", + "\u9676\u537f1" + ], + "affiliations": [ + "1. \u4e2d\u56fd\u4eba\u6c11\u89e3\u653e\u519b\u9646\u519b\u70ae\u5175\u9632\u7a7a\u5175\u5b66\u9662\u4fe1\u606f\u5de5\u7a0b\u7cfb", + "2. \u4e2d\u56fd\u4eba\u6c11\u89e3\u653e\u519b\u519b\u4e8b\u79d1\u5b66\u9662" + ], + "abstract": "\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u5177\u6709\u8106\u5f31\u6027\uff0c\u5bb9\u6613\u88ab\u7cbe\u5fc3\u8bbe\u8ba1\u7684\u5bf9\u6297\u6837\u672c\u653b\u51fb.\u68af\u5ea6\u653b\u51fb\u65b9\u6cd5\u5728\u767d\u76d2\u6a21\u578b\u4e0a\u653b\u51fb\u6210\u529f\u7387\u8f83\u9ad8\uff0c\u4f46\u5728\u9ed1\u76d2\u6a21\u578b\u4e0a\u7684\u8fc1\u79fb\u6027\u8f83\u5f31.\u57fa\u4e8eHeavy-ball\u578b\u52a8\u91cf\u548cNesterov\u578b\u52a8\u91cf\u7684\u68af\u5ea6\u653b\u51fb\u65b9\u6cd5\u7531\u4e8e\u5728\u66f4\u65b0\u65b9\u5411\u4e0a\u8003\u8651\u4e86\u5386\u53f2\u68af\u5ea6\u4fe1\u606f\uff0c\u63d0\u5347\u4e86\u5bf9\u6297\u6837\u672c\u7684\u8fc1\u79fb\u6027.\u4e3a\u4e86\u8fdb\u4e00\u6b65\u4f7f\u7528\u5386\u53f2\u68af\u5ea6\u4fe1\u606f\uff0c\u672c\u6587\u9488\u5bf9\u6536\u655b\u6027\u66f4\u597d\u7684Nesterov\u578b\u52a8\u91cf\u65b9\u6cd5\uff0c\u4f7f\u7528\u81ea\u9002\u5e94\u6b65\u957f\u7b56\u7565\u4ee3\u66ff\u76ee\u524d\u5e7f\u6cdb\u4f7f\u7528\u7684\u56fa\u5b9a\u6b65\u957f\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u65b9\u5411\u548c\u6b65\u957f\u5747\u4f7f\u7528\u5386\u53f2\u68af\u5ea6\u4fe1\u606f\u7684\u8fed\u4ee3\u5feb\u901f\u68af\u5ea6\u65b9\u6cd5\uff08Nesterov and Adaptive-learning-rate based Iterative Fast Gradient Method,NAI-FGM\uff09.\u6b64\u5916\uff0c\u672c\u6587\u8fd8\u63d0\u51fa\u4e86\u4e00\u79cd\u7ebf\u6027\u53d8\u6362\u4e0d\u53d8\u6027\uff08Linear-transformation Invariant Method,LIM\uff09\u7684\u6570\u636e\u589e\u5f3a\u65b9\u6cd5 .\u5b9e\u9a8c\u7ed3\u679c\u8bc1\u5b9e\u4e86NAI-FGM\u653b\u51fb\u65b9\u6cd5\u548cLIM\u6570\u636e\u589e\u5f3a\u7b56\u7565\u76f8\u5bf9\u4e8e\u540c\u7c7b\u578b\u65b9\u6cd5\u5747\u5177\u6709\u66f4\u9ad8\u7684\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387.\u7ec4\u5408NAI-FGM\u65b9\u6cd5\u548cLIM\u7b56\u7565\u751f\u6210\u5bf9\u6297\u6837\u672c\uff0c\u5728\u5e38\u89c4\u8bad\u7ec3\u6a21\u578b\u4e0a\u7684\u5e73\u5747\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387\u8fbe\u523087.8%\uff0c\u5728\u5bf9\u6297\u8bad\u7ec3\u6a21\u578b\u4e0a\u7684\u5e73\u5747\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387\u8fbe\u523057.5%\uff0c\u5728\u9632\u5fa1\u6a21\u578b\u4e0a\u7684\u5e73\u5747\u9ed1\u76d2\u653b\u51fb\u6210\u529f\u7387\u8fbe\u523067.2%\uff0c\u5747\u8d85\u8fc7\u73b0\u6709\u6700\u9ad8\u6c34\u5e73. " + }, + { + "name": "\u9488\u5bf9\u8eab\u4efd\u8bc1\u6587\u672c\u8bc6\u522b\u7684\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5\u7814\u7a76", + "authors": [ + "\u5f90\u660c\u51ef1,2", + "\u51af\u536b\u680b1,2", + "\u5f20\u6df3\u67701,2", + "\u90d1\u6653\u9f993,4,5", + "\u5f20\u8f896", + "\u738b\u98de\u8dc33,4,5" + ], + "affiliations": [ + "1. \u5317\u4eac\u4ea4\u901a\u5927\u5b66\u8ba1\u7b97\u673a\u4e0e\u4fe1\u606f\u6280\u672f\u5b66\u9662\u4fe1\u606f\u79d1\u5b66\u7814\u7a76\u6240", + "2. \u73b0\u4ee3\u4fe1\u606f\u79d1\u5b66\u4e0e\u7f51\u7edc\u6280\u672f\u5317\u4eac\u5e02\u91cd\u70b9\u5b9e\u9a8c\u5ba4", + "3. \u4e2d\u56fd\u79d1\u5b66\u9662\u81ea\u52a8\u5316\u7814\u7a76\u6240\u591a\u6a21\u6001\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u5168\u56fd\u91cd\u70b9\u5b9e\u9a8c\u5ba4", + "4. \u4e2d\u56fd\u79d1\u5b66\u9662\u81ea\u52a8\u5316\u7814\u7a76\u6240\u590d\u6742\u7cfb\u7edf\u7ba1\u7406\u4e0e\u63a7\u5236\u56fd\u5bb6\u91cd\u70b9\u5b9e\u9a8c\u5ba4", + "5. \u4e2d\u56fd\u79d1\u5b66\u9662\u5927\u5b66\u4eba\u5de5\u667a\u80fd\u5b66\u9662", + "6. \u5317\u4eac\u822a\u7a7a\u822a\u5929\u5927\u5b66\u4ea4\u901a\u79d1\u5b66\u4e0e\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u8eab\u4efd\u8bc1\u8ba4\u8bc1\u573a\u666f\u591a\u91c7\u7528\u6587\u672c\u8bc6\u522b\u6a21\u578b\u5bf9\u8eab\u4efd\u8bc1\u56fe\u7247\u7684\u5b57\u6bb5\u8fdb\u884c\u63d0\u53d6\u3001\u8bc6\u522b\u548c\u8eab\u4efd\u8ba4\u8bc1,\u5b58\u5728\u5f88\u5927\u7684\u9690\u79c1\u6cc4\u9732\u9690\u60a3.\u5e76\u4e14,\u5f53\u524d\u57fa\u4e8e\u6587\u672c\u8bc6\u522b\u6a21\u578b\u7684\u5bf9\u6297\u653b\u51fb\u7b97\u6cd5\u5927\u591a\u53ea\u8003\u8651\u7b80\u5355\u80cc\u666f\u7684\u6570\u636e(\u5982\u5370\u5237\u4f53)\u548c\u767d\u76d2\u6761\u4ef6,\u5f88\u96be\u5728\u7269\u7406\u4e16\u754c\u8fbe\u5230\u7406\u60f3\u7684\u653b\u51fb\u6548\u679c,\u4e0d\u9002\u7528\u4e8e\u590d\u6742\u80cc\u666f\u3001\u6570\u636e\u53ca\u9ed1\u76d2\u6761\u4ef6.\u4e3a\u7f13\u89e3\u4e0a\u8ff0\u95ee\u9898,\u672c\u6587\u63d0\u51fa\u9488\u5bf9\u8eab\u4efd\u8bc1\u6587\u672c\u8bc6\u522b\u6a21\u578b\u7684\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5,\u8003\u8651\u8f83\u4e3a\u590d\u6742\u7684\u56fe\u50cf\u80cc\u666f\u3001\u66f4\u4e25\u82db\u7684\u9ed1\u76d2\u6761\u4ef6\u4ee5\u53ca\u7269\u7406\u4e16\u754c\u7684\u653b\u51fb\u6548\u679c.\u672c\u7b97\u6cd5\u5728\u57fa\u4e8e\u8fc1\u79fb\u7684\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5\u7684\u57fa\u7840\u4e0a\u5f15\u5165\u4e8c\u503c\u5316\u63a9\u7801\u548c\u7a7a\u95f4\u53d8\u6362,\u5728\u4fdd\u8bc1\u653b\u51fb\u6210\u529f\u7387\u7684\u524d\u63d0\u4e0b\u63d0\u5347\u4e86\u5bf9\u6297\u6837\u672c\u7684\u89c6\u89c9\u6548\u679c\u548c\u7269\u7406\u4e16\u754c\u4e2d\u7684\u9c81\u68d2\u6027.\u901a\u8fc7\u63a2\u7d22\u4e0d\u540c\u8303\u6570\u9650\u5236\u4e0b\u57fa\u4e8e\u8fc1\u79fb\u7684\u9ed1\u76d2\u653b\u51fb\u7b97\u6cd5\u7684\u6027\u80fd\u4e0a\u9650\u548c\u5173\u952e\u8d85\u53c2\u6570\u7684\u5f71\u54cd,\u672c\u7b97\u6cd5\u5728\u767e\u5ea6\u8eab\u4efd\u8bc1\u8bc6\u522b\u6a21\u578b\u4e0a\u5b9e\u73b0\u4e86100%\u7684\u653b\u51fb\u6210\u529f\u7387.\u8eab\u4efd\u8bc1\u6570\u636e\u96c6\u540e\u7eed\u5c06\u5f00\u6e90." + }, + { + "name": "\u57fa\u4e8e\u667a\u80fd\u8fdb\u5316\u7b97\u6cd5\u7684\u53ef\u89c1\u6c34\u5370\u5bf9\u6297\u653b\u51fb", + "authors": [ + "\u5b63\u4fca\u8c6a1", + "\u5f20\u7389\u4e661", + "\u8d75\u82e5\u5b871", + "\u6e29\u6587\u5a962", + "\u8463\u74063" + ], + "affiliations": [ + "1. \u5357\u4eac\u822a\u7a7a\u822a\u5929\u5927\u5b66\u8ba1\u7b97\u673a\u79d1\u5b66\u4e0e\u6280\u672f\u5b66\u9662", + "2. \u6c5f\u897f\u8d22\u7ecf\u5927\u5b66\u4fe1\u606f\u7ba1\u7406\u5b66\u9662", + "3. \u5b81\u6ce2\u5927\u5b66\u4fe1\u606f\u79d1\u5b66\u4e0e\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u968f\u7740\u516c\u6c11\u7248\u6743\u610f\u8bc6\u7684\u63d0\u9ad8\uff0c\u8d8a\u6765\u8d8a\u591a\u542b\u6709\u6c34\u5370\u7684\u56fe\u50cf\u51fa\u73b0\u5728\u751f\u6d3b\u4e2d\u3002\u7136\u800c\uff0c\u73b0\u6709\u7684\u7814\u7a76\u8868\u660e\uff0c\u542b\u6709\u6c34\u5370\u7684\u56fe\u50cf\u4f1a\u5bfc\u81f4\u795e\u7ecf\u7f51\u7edc\u5206\u7c7b\u9519\u8bef\uff0c\u8fd9\u5bf9\u795e\u7ecf\u7f51\u7edc\u7684\u666e\u53ca\u548c\u5e94\u7528\u6784\u6210\u4e86\u5de8\u5927\u7684\u5a01\u80c1\u3002\u5bf9\u6297\u8bad\u7ec3\u662f\u89e3\u51b3\u8fd9\u7c7b\u95ee\u9898\u7684\u9632\u5fa1\u65b9\u6cd5\u4e4b\u4e00\uff0c\u4f46\u662f\u9700\u8981\u4f7f\u7528\u5927\u91cf\u7684\u6c34\u5370\u5bf9\u6297\u6837\u672c\u4f5c\u4e3a\u8bad\u7ec3\u6570\u636e\u3002\u4e3a\u6b64\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u667a\u80fd\u8fdb\u5316\u7b97\u6cd5\u7684\u53ef\u89c1\u6c34\u5370\u5bf9\u6297\u653b\u51fb\u65b9\u6cd5\u6765\u751f\u6210\u9ad8\u5f3a\u5ea6\u7684\u6c34\u5370\u5bf9\u6297\u6837\u672c\u3002\u8be5\u65b9\u6cd5\u4e0d\u4ec5\u80fd\u5feb\u901f\u751f\u6210\u6c34\u5370\u5bf9\u6297\u6837\u672c\uff0c\u800c\u4e14\u8fd8\u80fd\u4f7f\u5176\u6700\u5927\u7a0b\u5ea6\u5730\u653b\u51fb\u795e\u7ecf\u7f51\u7edc\u3002\u6b64\u5916\uff0c\u8be5\u65b9\u6cd5\u8fd8\u52a0\u5165\u4e86\u56fe\u50cf\u8d28\u91cf\u8bc4\u4ef7\u6307\u6807\u6765\u7ea6\u675f\u56fe\u50cf\u7684\u89c6\u89c9\u635f\u5931\uff0c\u4ece\u800c\u4f7f\u6c34\u5370\u5bf9\u6297\u6837\u672c\u66f4\u52a0\u7f8e\u89c2\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u6240\u63d0\u65b9\u6cd5\u76f8\u6bd4\u4e8e\u57fa\u51c6\u6c34\u5370\u653b\u51fb\u65b9\u6cd5\u65f6\u95f4\u590d\u6742\u5ea6\u66f4\u4f4e\uff0c\u76f8\u6bd4\u4e8e\u57fa\u51c6\u9ed1\u76d2\u653b\u51fb\u5bf9\u795e\u7ecf\u7f51\u7edc\u653b\u51fb\u6210\u529f\u7387\u66f4\u9ad8\u3002" + }, + { + "name": "\u57fa\u4e8e\u566a\u58f0\u7834\u574f\u548c\u6ce2\u5f62\u91cd\u5efa\u7684\u58f0\u7eb9\u5bf9\u6297\u6837\u672c\u9632\u5fa1\u65b9\u6cd5", + "authors": [ + "\u9b4f\u6625\u96e81", + "\u5b59\u84991", + "\u5f20\u96c4\u4f1f1", + "\u90b9\u971e1", + "\u5370\u67702" + ], + "affiliations": [ + "1. \u9646\u519b\u5de5\u7a0b\u5927\u5b66\u6307\u6325\u63a7\u5236\u5de5\u7a0b\u5b66\u9662", + "2. \u6c5f\u82cf\u8b66\u5b98\u5b66\u9662" + ], + "abstract": "\u8bed\u97f3\u662f\u4eba\u7c7b\u6700\u91cd\u8981\u7684\u4ea4\u6d41\u65b9\u5f0f\u4e4b\u4e00\u3002\u8bed\u97f3\u4fe1\u53f7\u4e2d\u9664\u4e86\u6587\u672c\u5185\u5bb9\u5916,\u8fd8\u5305\u542b\u4e86\u8bf4\u8bdd\u4eba\u7684\u8eab\u4efd\u3001\u79cd\u65cf\u3001\u5e74\u9f84\u3001\u6027\u522b\u548c\u60c5\u611f\u7b49\u4e30\u5bcc\u7684\u4fe1\u606f,\u5176\u4e2d\u8bf4\u8bdd\u4eba\u8eab\u4efd\u7684\u8bc6\u522b\u4e5f\u88ab\u79f0\u4e3a\u58f0\u7eb9\u8bc6\u522b,\u662f\u4e00\u79cd\u751f\u7269\u7279\u5f81\u8bc6\u522b\u6280\u672f\u3002\u58f0\u7eb9\u5177\u6709\u83b7\u53d6\u65b9\u4fbf\u3001\u5bb9\u6613\u4fdd\u5b58\u3001\u4f7f\u7528\u7b80\u5355\u7b49\u7279\u70b9,\u800c\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u7684\u8fdb\u6b65\u4e5f\u6781\u5927\u5730\u4fc3\u8fdb\u4e86\u8bc6\u522b\u51c6\u786e\u7387\u7684\u63d0\u5347,\u56e0\u6b64,\u58f0\u7eb9\u8bc6\u522b\u5df2\u88ab\u5e94\u7528\u4e8e\u667a\u6167\u91d1\u878d\u3001\u667a\u80fd\u5bb6\u5c45\u3001\u8bed\u97f3\u52a9\u624b\u548c\u53f8\u6cd5\u8c03\u67e5\u7b49\u9886\u57df\u3002\u53e6\u4e00\u65b9\u9762,\u9488\u5bf9\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u7684\u5bf9\u6297\u6837\u672c\u653b\u51fb\u53d7\u5230\u4e86\u5e7f\u6cdb\u5173\u6ce8,\u5728\u8f93\u5165\u4fe1\u53f7\u4e2d\u6dfb\u52a0\u4e0d\u53ef\u611f\u77e5\u7684\u5fae\u5c0f\u6270\u52a8\u5373\u53ef\u5bfc\u81f4\u6a21\u578b\u9884\u6d4b\u7ed3\u679c\u9519\u8bef\u3002\u5bf9\u6297\u6837\u672c\u7684\u51fa\u73b0\u5bf9\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u58f0\u7eb9\u8bc6\u522b\u4e5f\u5c06\u9020\u6210\u5de8\u5927\u7684\u5b89\u5168\u5a01\u80c1\u3002\u73b0\u6709\u58f0\u7eb9\u5bf9\u6297\u6837\u672c\u9632\u5fa1\u65b9\u6cd5\u4f1a\u4e0d\u540c\u7a0b\u5ea6\u5730\u5f71\u54cd\u6b63\u5e38\u6837\u672c\u7684\u8bc6\u522b,\u5e76\u4e14\u5c40\u9650\u4e8e\u7279\u5b9a\u7684\u653b\u51fb\u65b9\u6cd5\u6216\u8bc6\u522b\u6a21\u578b,\u9c81\u68d2\u6027\u8f83\u5dee\u3002\u4e3a\u4e86\u4f7f\u5bf9\u6297\u9632\u5fa1\u80fd\u591f\u517c\u987e\u7ea0\u6b63\u9519\u8bef\u8f93\u51fa\u548c\u51c6\u786e\u8bc6\u522b\u6b63\u5e38\u6837\u672c\u4e24\u4e2a\u65b9\u9762,\u672c\u6587\u63d0\u51fa\u4e00\u79cd\u201c\u7834\u574f+\u91cd\u5efa\u201d\u7684\u4e24\u9636\u6bb5\u5bf9\u6297\u6837\u672c\u9632\u5fa1\u65b9\u6cd5\u3002\u7b2c\u4e00\u9636\u6bb5,\u5728\u5bf9\u6297\u6837\u672c\u4e2d\u6dfb\u52a0\u5177\u6709\u4e00\u5b9a\u4fe1\u566a\u6bd4\u5e45\u5ea6\u9650\u5236\u7684\u9ad8\u65af\u767d\u566a\u58f0,\u7834\u574f\u5bf9\u6297\u6270\u52a8\u7684\u7ed3\u6784\u8fdb\u800c\u6d88\u9664\u6837\u672c\u7684\u5bf9\u6297\u6027\u3002\u7b2c\u4e8c\u9636\u6bb5,\u5229\u7528\u63d0\u51fa\u7684\u540d\u4e3aSCAT-Wave-U-Net\u7684\u8bed\u97f3\u589e\u5f3a\u6a21\u578b\u91cd\u5efa\u539f\u59cb\u8bed\u97f3\u6837\u672c,\u901a\u8fc7\u5728Wave-U-Net\u6a21\u578b\u7ed3\u6784\u4e2d\u5f15\u5165Transformer\u5168\u5c40\u591a\u5934\u81ea\u6ce8\u610f\u529b\u548c\u5c42\u95f4\u4ea4\u53c9\u6ce8\u610f\u529b\u673a\u5236,\u4f7f\u6539\u8fdb\u540e\u7684\u6a21\u578b\u66f4\u6709\u52a9\u4e8e\u9632\u5fa1\u58f0\u7eb9\u5bf9\u6297\u6837\u672c\u653b\u51fb\u3002\u5b9e\u9a8c\u8868\u660e,\u63d0\u51fa\u7684\u9632\u5fa1\u65b9\u6cd5\u4e0d\u4f9d\u8d56\u4e8e\u7279\u5b9a\u58f0\u7eb9\u8bc6\u522b\u7cfb\u7edf\u548c\u5bf9\u6297\u6837\u672c\u653b\u51fb\u65b9\u5f0f,\u5728\u4e24\u79cd\u5178\u578b\u7684\u58f0\u7eb9\u8bc6\u522b\u7cfb\u7edf\u4e0b\u5bf9\u591a\u79cd\u7c7b\u578b\u5bf9\u6297\u6837\u672c\u653b\u51fb\u7684\u9632\u5fa1\u6548\u679c\u5747\u4f18\u4e8e\u5176\u4ed6\u9884\u5904\u7406\u9632\u5fa1\u65b9\u6cd5\u3002 " + }, + { + "name": "\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u81ea\u7136\u8bed\u8a00\u5904\u7406\u653b\u9632\u7814\u7a76\u7efc\u8ff0", + "authors": [ + "\u9a6c\u751c", + "\u5f20\u56fd\u6881", + "\u90ed\u6653\u519b" + ], + "affiliations": [ + "\u897f\u85cf\u6c11\u65cf\u5927\u5b66\u4fe1\u606f\u5de5\u7a0b\u5b66\u9662" + ], + "abstract": "\u968f\u7740\u4eba\u5de5\u667a\u80fd\u7684\u53d1\u5c55\uff0c\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406\uff08NLP\uff09\u9886\u57df\u5df2\u7ecf\u53d6\u5f97\u4e86\u663e\u8457\u8fdb\u6b65\u3002\u7136\u800c\uff0cNLP\u6a21\u578b\u8fd8\u5b58\u5728\u5b89\u5168\u6027\u6f0f\u6d1e\u3002\u6587\u7ae0\u5206\u6790\u4e86\u6df1\u5ea6\u5b66\u4e60\u5728NLP\u4e09\u5927\u6838\u5fc3\u4efb\u52a1\uff08\u5305\u62ec\u6587\u672c\u8868\u793a\u3001\u8bed\u5e8f\u5efa\u6a21\u548c\u77e5\u8bc6\u8868\u793a\uff09\u4e2d\u7684\u5e94\u7528\u73b0\u72b6\uff0c\u9488\u5bf9\u6587\u672c\u751f\u6210\u3001\u6587\u672c\u5206\u7c7b\u4ee5\u53ca\u8bed\u4e49\u89e3\u6790\u9762\u4e34\u7684\u653b\u51fb\u6280\u672f\uff0c\u63a2\u8ba8\u4e86\u5bf9\u6297\u6027\u8bad\u7ec3\u3001\u6b63\u5219\u5316\u6280\u672f\u3001\u6a21\u578b\u84b8\u998f\u7b49\u4e00\u7cfb\u5217\u9632\u5fa1\u6280\u672f\u5728\u5b9e\u9645\u5e94\u7528\u4e2d\u7684\u6548\u7528\u548c\u5c40\u9650\uff0c\u5e76\u901a\u8fc7\u6587\u672c\u5206\u7c7b\u4efb\u52a1\u7684\u5b9e\u8bc1\u7814\u7a76\u9a8c\u8bc1\u4e86\u96c6\u6210\u5bf9\u6297\u8bad\u7ec3\u7684\u6709\u6548\u6027\u3002" + }, + { + "name": "\u4e00\u79cd\u57fa\u4e8e\u8f6e\u5ed3\u7a00\u758f\u5bf9\u6297\u7684\u89c6\u9891\u6b65\u6001\u9690\u79c1\u4fdd\u62a4\u7b97\u6cd5", + "authors": [ + "\u8bb8\u53ef", + "\u674e\u5609\u6021", + "\u848b\u5174\u6d69", + "\u5b59\u952c\u950b" + ], + "affiliations": [ + "\u4e0a\u6d77\u4ea4\u901a\u5927\u5b66\u7f51\u7edc\u7a7a\u95f4\u5b89\u5168\u5b66\u9662" + ], + "abstract": "\u6df1\u5ea6\u7f51\u7edc\u6a21\u578b\u53ef\u4ee5\u4ece\u89c6\u9891\u6b65\u6001\u5e8f\u5217\u4e2d\u83b7\u53d6\u4eba\u4f53\u6b65\u6001\u751f\u7269\u7279\u5f81\u5e76\u8bc6\u522b\u4eba\u7269\u8eab\u4efd,\u9020\u6210\u4e25\u91cd\u7684\u9690\u79c1\u6cc4\u9732\u5b89\u5168\u5a01\u80c1\u3002\u73b0\u6709\u65b9\u6cd5\u4e00\u822c\u901a\u8fc7\u5bf9\u89c6\u9891\u753b\u9762\u4e2d\u7684\u4eba\u4f53\u8fdb\u884c\u6a21\u7cca\u3001\u53d8\u5f62\u7b49\u5904\u7406\u6765\u4fdd\u62a4\u9690\u79c1,\u8fd9\u4e9b\u65b9\u6cd5\u53ef\u4ee5\u5728\u4e00\u5b9a\u7a0b\u5ea6\u4e0a\u6539\u53d8\u4eba\u4f53\u5916\u89c2,\u4f46\u5f88\u96be\u6539\u53d8\u4eba\u7269\u884c\u8d70\u59ff\u6001,\u96be\u4ee5\u9003\u907f\u6df1\u5ea6\u7f51\u7edc\u6a21\u578b\u7684\u8bc6\u522b,\u4e14\u8fd9\u79cd\u5904\u7406\u5f80\u5f80\u4f34\u968f\u7740\u5bf9\u89c6\u9891\u8d28\u91cf\u7684\u4e25\u91cd\u7834\u574f,\u964d\u4f4e\u4e86\u89c6\u9891\u7684\u89c6\u89c9\u53ef\u7528\u6027\u3002\u9488\u5bf9\u8be5\u95ee\u9898,\u6587\u7ae0\u63d0\u51fa\u4e00\u79cd\u57fa\u4e8e\u8f6e\u5ed3\u7a00\u758f\u5bf9\u6297\u7684\u89c6\u9891\u6b65\u6001\u9690\u79c1\u4fdd\u62a4\u7b97\u6cd5,\u901a\u8fc7\u5bf9\u6b65\u6001\u8bc6\u522b\u6a21\u578b\u7684\u5bf9\u6297\u653b\u51fb\u6765\u8ba1\u7b97\u753b\u9762\u4e2d\u4eba\u4f53\u8f6e\u5ed3\u5468\u56f4\u7684\u6709\u6548\u4fee\u6539\u4f4d\u7f6e\u3002\u4e0e\u4f20\u7edf\u65b9\u6cd5\u76f8\u6bd4,\u5728\u5177\u6709\u76f8\u540c\u9690\u79c1\u4fdd\u62a4\u80fd\u529b\u7684\u60c5\u51b5\u4e0b,\u8be5\u7b97\u6cd5\u51cf\u5c11\u4e86\u5bf9\u753b\u9762\u7684\u4fee\u6539,\u5728\u9690\u79c1\u5b89\u5168\u6027\u548c\u89c6\u89c9\u53ef\u7528\u6027\u4e0a\u8fbe\u5230\u4e86\u8f83\u597d\u7684\u5747\u8861\u3002\u8be5\u7b97\u6cd5\u5728\u516c\u5f00\u6b65\u6001\u6570\u636e\u5e93CASIA-B\u548cOUMVLP\u4e0a\u5bf94\u79cd\u6b65\u6001\u8bc6\u522b\u6a21\u578b\u8fdb\u884c\u6d4b\u8bd5,\u901a\u8fc7\u4e0e\u4e0d\u540c\u6b65\u6001\u9690\u79c1\u4fdd\u62a4\u65b9\u6cd5\u5bf9\u6bd4,\u9a8c\u8bc1\u4e86\u8be5\u7b97\u6cd5\u5728\u6b65\u6001\u9690\u79c1\u4fdd\u62a4\u4e0a\u7684\u6709\u6548\u6027\u548c\u53ef\u7528\u6027\u3002" + } +] \ No newline at end of file diff --git a/result_arxiv_knowledge_graph.json b/result_arxiv_knowledge_graph.json new file mode 100644 index 0000000..7f5055f --- /dev/null +++ b/result_arxiv_knowledge_graph.json @@ -0,0 +1 @@ +[{"name": "Solving Power Grid Optimization Problems with Rydberg Atoms", "authors": "Nora Bauer,K\u00fcbra Yeter-Aydeniz,Elias Kokkas,George Siopsis", "affiliations": "no", "abstract": "The rapid development of neutral atom quantum hardware provides a unique opportunity to design hardware-centered algorithms for solving real-world problems aimed at establishing quantum utility. In this work, we study the performance of two such algorithms on solving MaxCut problem for various weighted graphs. The first method uses a state-of-the-art machine learning tool to optimize the pulse shape and embedding of the graph using an adiabatic Ansatz to find the ground state. We tested the performance of this method on finding maximum power section task of the IEEE 9-bus power system and obtaining MaxCut of randomly generated problems of size up to 12 on the Aquila quantum processor. To the best of our knowledge, this work presents the first MaxCut results on Quera's Aquila quantum hardware. Our experiments run on Aquila demonstrate that even though the probability of obtaining the solution is reduced, one can still solve the MaxCut problem on cloud-accessed neutral atom quantum hardware. The second method uses local detuning, which is an emergent update on the Aquila hardware, to obtain a near exact realization of the standard QAOA Ansatz with similar performance. Finally, we study the fidelity throughout the time evolution realized in the adiabatic method as a benchmark for the IEEE 9-bus power grid graph state."}, {"name": "Towards Human Awareness in Robot Task Planning with Large Language Models", "authors": "Yuchen Liu,Luigi Palmieri,Sebastian Koch,Ilche Georgievski,Marco Aiello", "affiliations": "no", "abstract": "The recent breakthroughs in the research on Large Language Models (LLMs) have triggered a transformation across several research domains. Notably, the integration of LLMs has greatly enhanced performance in robot Task And Motion Planning (TAMP). However, previous approaches often neglect the consideration of dynamic environments, i.e., the presence of dynamic objects such as humans. In this paper, we propose a novel approach to address this gap by incorporating human awareness into LLM-based robot task planning. To obtain an effective representation of the dynamic environment, our approach integrates humans' information into a hierarchical scene graph. To ensure the plan's executability, we leverage LLMs to ground the environmental topology and actionable knowledge into formal planning language. Most importantly, we use LLMs to predict future human activities and plan tasks for the robot considering the predictions. Our contribution facilitates the development of integrating human awareness into LLM-driven robot task planning, and paves the way for proactive robot decision-making in dynamic environments."}, {"name": "EEG_GLT-Net: Optimising EEG Graphs for Real-time Motor Imagery Signals Classification", "authors": "Htoo Wai Aung,Jiao Jiao Li,Yang An,Steven W. Su", "affiliations": "no", "abstract": "Brain-Computer Interfaces connect the brain to external control devices, necessitating the accurate translation of brain signals such as from electroencephalography (EEG) into executable commands. Graph Neural Networks (GCN) have been increasingly applied for classifying EEG Motor Imagery signals, primarily because they incorporates the spatial relationships among EEG channels, resulting in improved accuracy over traditional convolutional methods. Recent advances by GCNs-Net in real-time EEG MI signal classification utilised Pearson Coefficient Correlation (PCC) for constructing adjacency matrices, yielding significant results on the PhysioNet dataset. Our paper introduces the EEG Graph Lottery Ticket (EEG_GLT) algorithm, an innovative technique for constructing adjacency matrices for EEG channels. It does not require pre-existing knowledge of inter-channel relationships, and it can be tailored to suit both individual subjects and GCN model architectures. Our findings demonstrated that the PCC method outperformed the Geodesic approach by 9.65% in mean accuracy, while our EEG_GLT matrix consistently exceeded the performance of the PCC method by a mean accuracy of 13.39%. Also, we found that the construction of the adjacency matrix significantly influenced accuracy, to a greater extent than GCN model configurations. A basic GCN configuration utilising our EEG_GLT matrix exceeded the performance of even the most complex GCN setup with a PCC matrix in average accuracy. Our EEG_GLT method also reduced MACs by up to 97% compared to the PCC method, while maintaining or enhancing accuracy. In conclusion, the EEG_GLT algorithm marks a breakthrough in the development of optimal adjacency matrices, effectively boosting both computational accuracy and efficiency, making it well-suited for real-time classification of EEG MI signals that demand intensive computational resources."}, {"name": "Graph Continual Learning with Debiased Lossless Memory Replay", "authors": "Chaoxi Niu,Guansong Pang,Ling Chen", "affiliations": "no", "abstract": "Real-life graph data often expands continually, rendering the learning of graph neural networks (GNNs) on static graph data impractical. Graph continual learning (GCL) tackles this problem by continually adapting GNNs to the expanded graph of the current task while maintaining the performance over the graph of previous tasks. Memory replay-based methods, which aim to replay data of previous tasks when learning new tasks, have been explored as one principled approach to mitigate the forgetting of the knowledge learned from the previous tasks. In this paper we extend this methodology with a novel framework, called Debiased Lossless Memory replay (DeLoMe). Unlike existing methods that sample nodes/edges of previous graphs to construct the memory, DeLoMe learns small lossless synthetic node representations as the memory. The learned memory can not only preserve the graph data privacy but also capture the holistic graph information, for which the sampling-based methods are not viable. Further, prior methods suffer from bias toward the current task due to the data imbalance between the classes in the memory data and the current data. A debiased GCL loss function is devised in DeLoMe to effectively alleviate this bias. Extensive experiments on four graph datasets show the effectiveness of DeLoMe under both class- and task-incremental learning settings."}, {"name": "Neuromorphic Vision-based Motion Segmentation with Graph Transformer Neural Network", "authors": "Yusra Alkendi,Rana Azzam,Sajid Javed,Lakmal Seneviratne,Yahya Zweiri", "affiliations": "no", "abstract": "Moving object segmentation is critical to interpret scene dynamics for robotic navigation systems in challenging environments. Neuromorphic vision sensors are tailored for motion perception due to their asynchronous nature, high temporal resolution, and reduced power consumption. However, their unconventional output requires novel perception paradigms to leverage their spatially sparse and temporally dense nature. In this work, we propose a novel event-based motion segmentation algorithm using a Graph Transformer Neural Network, dubbed GTNN. Our proposed algorithm processes event streams as 3D graphs by a series of nonlinear transformations to unveil local and global spatiotemporal correlations between events. Based on these correlations, events belonging to moving objects are segmented from the background without prior knowledge of the dynamic scene geometry. The algorithm is trained on publicly available datasets including MOD, EV-IMO, and \\textcolor{black}{EV-IMO2} using the proposed training scheme to facilitate efficient training on extensive datasets. Moreover, we introduce the Dynamic Object Mask-aware Event Labeling (DOMEL) approach for generating approximate ground-truth labels for event-based motion segmentation datasets. We use DOMEL to label our own recorded Event dataset for Motion Segmentation (EMS-DOMEL), which we release to the public for further research and benchmarking. Rigorous experiments are conducted on several unseen publicly-available datasets where the results revealed that GTNN outperforms state-of-the-art methods in the presence of dynamic background variations, motion patterns, and multiple dynamic objects with varying sizes and velocities. GTNN achieves significant performance gains with an average increase of 9.4% and 4.5% in terms of motion segmentation accuracy (IoU%) and detection rate (DR%), respectively."}, {"name": "Classical and Quantum Distributed Algorithms for the Survivable Network Design Problem", "authors": "Phillip Kerger,David E. Bernal Neira,Zoe Gonzalez Izquierdo,Eleanor G. Rieffel", "affiliations": "no", "abstract": "We investigate distributed classical and quantum approaches for the survivable network design problem (SNDP), sometimes called the generalized Steiner problem. These problems generalize many complex graph problems of interest, such as the traveling salesperson problem, the Steiner tree problem, and the k-connected network problem. To our knowledge, no classical or quantum algorithms for the SNDP have been formulated in the distributed settings we consider. We describe algorithms that are heuristics for the general problem but give concrete approximation bounds under specific parameterizations of the SNDP, which in particular hold for the three aforementioned problems that SNDP generalizes. We use a classical, centralized algorithmic framework first studied in (Goemans & Bertsimas 1993) and provide a distributed implementation thereof. Notably, we obtain asymptotic quantum speedups by leveraging quantum shortest path computations in this framework, generalizing recent work of (Kerger et al. 2023). These results raise the question of whether there is a separation between the classical and quantum models for application-scale instances of the problems considered."}] \ No newline at end of file diff --git a/t1.py b/t1.py new file mode 100644 index 0000000..71cab14 --- /dev/null +++ b/t1.py @@ -0,0 +1,14 @@ +from serpapi import GoogleSearch +# GoogleSearch +params = { + "q": "Coffee", + "location": "Austin, Texas, United States", + "hl": "en", + "gl": "us", + "google_domain": "google.com", + "api_key": "681ac1d6fe9958124d39f25ea5afd759b63f45e52cac7e85629655024661166e" +} + +search = GoogleSearch(params) +results = search.get_dict() +print(results) diff --git a/te_u/arxiv.py b/te_u/arxiv.py new file mode 100644 index 0000000..4161211 --- /dev/null +++ b/te_u/arxiv.py @@ -0,0 +1,150 @@ +import undetected_chromedriver as uc +import time +import random +import json +import matplotlib.pyplot as plt # 数据可视化 +import jieba # 词语切割 +import wordcloud # 分词 +from wordcloud import WordCloud, ImageColorGenerator, STOPWORDS # 词云，颜色生成器，停止词 +import numpy as np # 科学计算 +from PIL import Image # 处理图片 +from bs4 import BeautifulSoup +from lxml import etree + + +# def get_current_page_result(driver): +# """ 采集一页里的所有item """ +# result_area = driver.find_element(by="id", value="ModuleSearchResult") +# current_page_results = result_area.find_elements(by="xpath", value='//tbody/tr') +# +# names = [r.find_element(by="xpath", value='td[@class="name"]') for r in current_page_results] +# links = [r.find_element(by="xpath", value='td[@class="name"]/a').get_attribute("href") for r in current_page_results] +# +# items = get_items(driver, links) +# return items + + +def get_items(driver, links): + items = [] + for i, l in enumerate(links): + item = get_item(driver, l) + items.append(item) + return items + + +def get_item(driver, link): + item = {} + driver.get(link) # 获取新的论文链接 + time.sleep(5 + 3 * random.random()) # 等等加载完成 + + # 标题 + name = driver.find_element(by="xpath", value='//h1[contains(@class, "title")]').text + item["name"] = name + + # 作者 + names_element = driver.find_elements(by="xpath", value='//div[@class="authors"]//a') + names = [n_ele.text for n_ele in names_element] + item["authors"] = ",".join(names) + + # 单位 + item["affiliations"] = "no" + + # 摘要 + # 如果有更多，先点更多 + # try: + # more_bn = driver.find_element(by="id", value="ChDivSummaryMore") + # more_bn.click() + # time.sleep(1 + 1 * random.random()) # 等等加载完成 + # except: + # more_bn = None + + abstract_area = driver.find_element(by="xpath", value='//blockquote[contains(@class, "abstract")]') + abstract = abstract_area.text + item["abstract"] = abstract + + return item + + +def get_links_etree(driver): + dom = etree.HTML(driver.page_source) + links = dom.xpath('//ol[@class="breathe-horizontal"]/li/div/p/a/@href') + return links + + +def get_news_from_arxiv(total_num, keyword): + keyword = [i.strip() for i in keyword.strip().split()] + url = f"https://arxiv.org/search/?query={'+'.join(keyword)}&searchtype=all&source=header" + driver = uc.Chrome() + driver.get(url) + # time.sleep(3 + 2 * random.random()) # 等等加载完成 + # # 搜索 + # input_button = driver.find_element(by="id", value="txt_SearchText") + # input_button.send_keys(keyword) + # time.sleep(1 + 1 * random.random()) # 等等加载完成 + # + # search_bn = driver.find_element(by="xpath", value='//input[@class="search-btn"]') + # search_bn.click() + time.sleep(5 + 3 * random.random()) # 等等加载完成 + + # 获取相应的链接 + links = [] + stop_flag = False + + while not stop_flag: + link_current_page = get_links_etree(driver) + links.extend(link_current_page) + + if len(links) < total_num: + # 下一页 + try: + next_page_btn = driver.find_element(by="xpath", value='//a[@class="pagination-next"]') + next_page_btn.click() + time.sleep(2 + 2 * random.random()) # 等等加载完成 + # driver.refresh() + # time.sleep(2 + 2 * random.random()) # 等等加载完成 + except Exception as e: + print("没有下一页，返回当前的采集的所有结果", e) + stop_flag = True + total_num = len(links) + else: + # 超过了需要的连接数就停止 + stop_flag = True + + links = links[:total_num] + + results = get_items(driver, links) + + with open(f"result_arxiv_{'_'.join(keyword)}.json", "w", encoding="utf8") as f: + f.write(json.dumps(results)) + + driver.close() + return results + + +def get_clouds(word_list): + text = "，".join(word_list) + wordlist = jieba.lcut(text) # 切割词语 + space_list = ' '.join(wordlist) # 空格链接词语 + # backgroud = np.array(Image.open('test1.jpg')) + + wc = WordCloud(width=400, height=300, + background_color='white', + mode='RGB', + # mask=backgroud, # 添加蒙版，生成指定形状的词云，并且词云图的颜色可从蒙版里提取 + max_words=200, + stopwords=STOPWORDS.update(('in', "of", "for")), # 内置的屏蔽词,并添加自己设置的词语 + font_path='C:\Windows\Fonts\STZHONGS.ttf', + max_font_size=100, + relative_scaling=0.6, # 设置字体大小与词频的关联程度为0.4 + random_state=50, + scale=2 + ).generate(space_list) + + # image_color = ImageColorGenerator(backgroud) # 设置生成词云的颜色，如去掉这两行则字体为默认颜色 + # wc.recolor(color_func=image_color) + + return wc.to_array() + + +if __name__ == '__main__': + get_news_from_arxiv(5, "knowledge graph") diff --git a/te_u/paper_down_load/csv/ECCV_2022.csv b/te_u/paper_down_load/csv/ECCV_2022.csv new file mode 100644 index 0000000..5b95b71 --- /dev/null +++ b/te_u/paper_down_load/csv/ECCV_2022.csv @@ -0,0 +1,1646 @@ +title,main link,supplemental link +learning-depth-from-focus-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610001-supp.pdf +learning-based-point-cloud-registration-for-6d-object-pose-estimation-in-the-real-world,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610018.pdf, +an-end-to-end-transformer-model-for-crowd-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610037.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610037-supp.pdf +few-shot-single-view-3d-reconstruction-with-memory-prior-contrastive-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610054.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610054-supp.pdf +did-m3d-decoupling-instance-depth-for-monocular-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610071.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610071-supp.pdf +adaptive-co-teaching-for-unsupervised-monocular-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610089.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610089-supp.pdf +fusing-local-similarities-for-retrieval-based-3d-orientation-estimation-of-unseen-objects,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610106.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610106-supp.pdf +lidar-point-cloud-guided-monocular-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610123.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610123-supp.pdf +structural-causal-3d-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610140.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610140-supp.pdf +3d-human-pose-estimation-using-mobius-graph-convolutional-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610158.pdf, +learning-to-train-a-point-cloud-reconstruction-network-without-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610177.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610177-supp.pdf +panoformer-panorama-transformer-for-indoor-360deg-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610193.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610193-supp.pdf +self-supervised-human-mesh-recovery-with-cross-representation-alignment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610210.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610210-supp.pdf +alignsdf-pose-aligned-signed-distance-fields-for-hand-object-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610229.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610229-supp.zip +a-reliable-online-method-for-joint-estimation-of-focal-length-and-camera-rotation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610247.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610247-supp.pdf +ps-nerf-neural-inverse-rendering-for-multi-view-photometric-stereo,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610263.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610263-supp.pdf +share-with-thy-neighbors-single-view-reconstruction-by-cross-instance-consistency,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610282.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610282-supp.pdf +towards-comprehensive-representation-enhancement-in-semantics-guided-self-supervised-monocular-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610299.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610299-supp.zip +avatarcap-animatable-avatar-conditioned-monocular-human-volumetric-capture,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610317.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610317-supp.pdf +cross-attention-of-disentangled-modalities-for-3d-human-mesh-recovery-with-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610336.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610336-supp.pdf +georefine-self-supervised-online-depth-refinement-for-accurate-dense-mapping,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610354.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610354-supp.pdf +multi-modal-masked-pre-training-for-monocular-panoramic-depth-completion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610372.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610372-supp.pdf +gitnet-geometric-prior-based-transformation-for-birds-eye-view-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610390.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610390-supp.pdf +learning-visibility-for-robust-dense-human-body-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610406.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610406-supp.pdf +towards-high-fidelity-single-view-holistic-reconstruction-of-indoor-scenes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610423.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610423-supp.pdf +compnvs-novel-view-synthesis-with-scene-completion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610441.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610441-supp.pdf +sketchsampler-sketch-based-3d-reconstruction-via-view-dependent-depth-sampling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610457.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610457-supp.pdf +localbins-improving-depth-estimation-by-learning-local-distributions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610473.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610473-supp.pdf +2d-gans-meet-unsupervised-single-view-3d-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610490.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610490-supp.pdf +infinitenature-zero-learning-perpetual-view-generation-of-natural-scenes-from-single-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610508.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610508-supp.pdf +semi-supervised-single-view-3d-reconstruction-via-prototype-shape-priors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610528.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610528-supp.pdf +bilateral-normal-integration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610545.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610545-supp.pdf +s2contact-graph-based-network-for-3d-hand-object-contact-estimation-with-semi-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610561.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610561-supp.pdf +sc-wls-towards-interpretable-feed-forward-camera-re-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610578.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610578-supp.pdf +floatingfusion-depth-from-tof-and-image-stabilized-stereo-cameras,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610595.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610595-supp.pdf +deltar-depth-estimation-from-a-light-weight-tof-sensor-and-rgb-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610612.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610612-supp.zip +3d-room-layout-estimation-from-a-cubemap-of-panorama-image-via-deep-manhattan-hough-transform,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610630.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610630-supp.pdf +rbp-pose-residual-bounding-box-projection-for-category-level-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610647.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610647-supp.pdf +monocular-3d-object-reconstruction-with-gan-inversion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610665.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610665-supp.pdf +map-free-visual-relocalization-metric-pose-relative-to-a-single-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610682.pdf, +self-distilled-feature-aggregation-for-self-supervised-monocular-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610700.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610700-supp.pdf +planes-vs-chairs-category-guided-3d-shape-learning-without-any-3d-cues,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610717.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136610717-supp.pdf +mhr-net-multiple-hypothesis-reconstruction-of-non-rigid-shapes-from-2d-views,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620001-supp.pdf +depth-map-decomposition-for-monocular-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620018.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620018-supp.pdf +monitored-distillation-for-positive-congruent-depth-completion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620035.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620035-supp.pdf +resolution-free-point-cloud-sampling-network-with-data-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620053.pdf, +organic-priors-in-non-rigid-structure-from-motion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620069.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620069-supp.pdf +perspective-flow-aggregation-for-data-limited-6d-object-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620087.pdf, +danbo-disentangled-articulated-neural-body-representations-via-graph-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620104.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620104-supp.pdf +chore-contact-human-and-object-reconstruction-from-a-single-rgb-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620121.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620121-supp.pdf +learned-vertex-descent-a-new-direction-for-3d-human-model-fitting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620141.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620141-supp.pdf +self-calibrating-photometric-stereo-by-neural-inverse-rendering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620160.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620160-supp.pdf +3d-clothed-human-reconstruction-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620177.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620177-supp.pdf +directed-ray-distance-functions-for-3d-scene-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620193.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620193-supp.pdf +object-level-depth-reconstruction-for-category-level-6d-object-pose-estimation-from-monocular-rgb-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620212.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620212-supp.pdf +uncertainty-quantification-in-depth-estimation-via-constrained-ordinal-regression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620229.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620229-supp.pdf +costdcnet-cost-volume-based-depth-completion-for-a-single-rgb-d-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620248.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620248-supp.pdf +shapo-implicit-representations-for-multi-object-shape-appearance-and-pose-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620266.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620266-supp.zip +3d-siamese-transformer-network-for-single-object-tracking-on-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620284.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620284-supp.pdf +object-wake-up-3d-object-rigging-from-a-single-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620302.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620302-supp.pdf +integratedpifu-integrated-pixel-aligned-implicit-function-for-single-view-human-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620319.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620319-supp.pdf +realistic-one-shot-mesh-based-head-avatars,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620336.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620336-supp.pdf +a-kendall-shape-space-approach-to-3d-shape-estimation-from-2d-landmarks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620354.pdf, +neural-light-field-estimation-for-street-scenes-with-differentiable-virtual-object-insertion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620370.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620370-supp.pdf +perspective-phase-angle-model-for-polarimetric-3d-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620387.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620387-supp.zip +deepshadow-neural-shape-from-shadow,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620403.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620403-supp.pdf +camera-auto-calibration-from-the-steiner-conic-of-the-fundamental-matrix,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620419.pdf, +super-resolution-3d-human-shape-from-a-single-low-resolution-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620435.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620435-supp.pdf +minimal-neural-atlas-parameterizing-complex-surfaces-with-minimal-charts-and-distortion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620452.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620452-supp.pdf +extrudenet-unsupervised-inverse-sketch-and-extrude-for-shape-parsing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620468.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620468-supp.pdf +catre-iterative-point-clouds-alignment-for-category-level-object-pose-refinement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620485.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620485-supp.pdf +optimization-over-disentangled-encoding-unsupervised-cross-domain-point-cloud-completion-via-occlusion-factor-manipulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620504.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620504-supp.zip +unsupervised-learning-of-3d-semantic-keypoints-with-mutual-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620521.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620521-supp.pdf +mvdecor-multi-view-dense-correspondence-learning-for-fine-grained-3d-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620538.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620538-supp.pdf +supr-a-sparse-unified-part-based-human-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620555.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620555-supp.pdf +revisiting-point-cloud-simplification-a-learnable-feature-preserving-approach,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620573.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620573-supp.pdf +masked-autoencoders-for-point-cloud-self-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620591.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620591-supp.pdf +intrinsic-neural-fields-learning-functions-on-manifolds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620609.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620609-supp.zip +skeleton-free-pose-transfer-for-stylized-3d-characters,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620627.pdf, +masked-discrimination-for-self-supervised-learning-on-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620645.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620645-supp.pdf +fbnet-feedback-network-for-point-cloud-completion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620664.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620664-supp.pdf +meta-sampler-almost-universal-yet-task-oriented-sampling-for-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620682.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620682-supp.pdf +a-level-set-theory-for-neural-implicit-evolution-under-explicit-flows,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620699.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620699-supp.pdf +efficient-point-cloud-analysis-using-hilbert-curve,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620717.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136620717-supp.pdf +toch-spatio-temporal-object-to-hand-correspondence-for-motion-refinement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630001-supp.zip +laterf-label-and-text-driven-object-radiance-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630021.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630021-supp.pdf +meshmae-masked-autoencoders-for-3d-mesh-data-analysis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630038.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630038-supp.pdf +unsupervised-deep-multi-shape-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630056.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630056-supp.pdf +texturify-generating-textures-on-3d-shape-surfaces,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630073.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630073-supp.zip +autoregressive-3d-shape-generation-via-canonical-mapping,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630091.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630091-supp.pdf +pointtree-transformation-robust-point-cloud-encoder-with-relaxed-k-d-trees,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630107.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630107-supp.pdf +unif-united-neural-implicit-functions-for-clothed-human-reconstruction-and-animation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630123.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630123-supp.pdf +prif-primary-ray-based-implicit-function,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630140.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630140-supp.pdf +point-cloud-domain-adaptation-via-masked-local-3d-structure-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630159.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630159-supp.pdf +clip-actor-text-driven-recommendation-and-stylization-for-animating-human-meshes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630176.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630176-supp.pdf +planeformers-from-sparse-view-planes-to-3d-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630194.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630194-supp.pdf +learning-implicit-templates-for-point-based-clothed-human-modeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630211.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630211-supp.zip +exploring-the-devil-in-graph-spectral-domain-for-3d-point-cloud-attacks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630230.pdf, +structure-aware-editable-morphable-model-for-3d-facial-detail-animation-and-manipulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630248.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630248-supp.zip +mofanerf-morphable-facial-neural-radiance-field,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630267.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630267-supp.zip +pointinst3d-segmenting-3d-instances-by-points,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630284.pdf, +cross-modal-3d-shape-generation-and-manipulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630300.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630300-supp.pdf +latent-partition-implicit-with-surface-codes-for-3d-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630318.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630318-supp.pdf +implicit-field-supervision-for-robust-non-rigid-shape-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630338.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630338-supp.pdf +learning-self-prior-for-mesh-denoising-using-dual-graph-convolutional-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630358.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630358-supp.pdf +diffconv-analyzing-irregular-point-clouds-with-an-irregular-view,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630375.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630375-supp.zip +pd-flow-a-point-cloud-denoising-framework-with-normalizing-flows,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630392.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630392-supp.pdf +seedformer-patch-seeds-based-point-cloud-completion-with-upsample-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630409.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630409-supp.pdf +deepmend-learning-occupancy-functions-to-represent-shape-for-repair,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630426.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630426-supp.pdf +a-repulsive-force-unit-for-garment-collision-handling-in-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630444.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630444-supp.pdf +shape-pose-disentanglement-using-se-3-equivariant-vector-neurons,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630461.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630461-supp.zip +3d-equivariant-graph-implicit-functions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630477.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630477-supp.pdf +patchrd-detail-preserving-shape-completion-by-learning-patch-retrieval-and-deformation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630494.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630494-supp.pdf +3d-shape-sequence-of-human-comparison-and-classification-using-current-and-varifolds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630514.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630514-supp.zip +conditional-flow-nerf-accurate-3d-modelling-with-reliable-uncertainty-quantification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630531.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630531-supp.zip +unsupervised-pose-aware-part-decomposition-for-man-made-articulated-objects,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630549.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630549-supp.pdf +meshudf-fast-and-differentiable-meshing-of-unsigned-distance-field-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630566.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630566-supp.pdf +spe-net-boosting-point-cloud-analysis-via-rotation-robustness-enhancement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630582.pdf, +the-shape-part-slot-machine-contact-based-reasoning-for-generating-3d-shapes-from-parts,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630599.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630599-supp.pdf +spatiotemporal-self-attention-modeling-with-temporal-patch-shift-for-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630615.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630615-supp.pdf +proposal-free-temporal-action-detection-via-global-segmentation-mask-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630632.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630632-supp.pdf +semi-supervised-temporal-action-detection-with-proposal-free-masking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630649.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630649-supp.pdf +zero-shot-temporal-action-detection-via-vision-language-prompting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630667.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630667-supp.pdf +cycda-unsupervised-cycle-domain-adaptation-to-learn-from-image-to-video,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630684.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630684-supp.pdf +s2n-suppression-strengthen-network-for-event-based-recognition-under-variant-illuminations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630701.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630701-supp.pdf +cmd-self-supervised-3d-action-representation-learning-with-cross-modal-mutual-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630719.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136630719-supp.pdf +expanding-language-image-pretrained-models-for-general-video-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640001-supp.pdf +hunting-group-clues-with-transformers-for-social-group-activity-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640018.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640018-supp.pdf +contrastive-positive-mining-for-unsupervised-3d-action-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640035.pdf, +target-absent-human-attention,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640051.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640051-supp.pdf +uncertainty-based-spatial-temporal-attention-for-online-action-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640068.pdf, +iwin-human-object-interaction-detection-via-transformer-with-irregular-windows,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640085.pdf, +rethinking-zero-shot-action-recognition-learning-from-latent-atomic-actions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640102.pdf, +mining-cross-person-cues-for-body-part-interactiveness-learning-in-hoi-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640119.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640119-supp.pdf +collaborating-domain-shared-and-target-specific-feature-clustering-for-cross-domain-3d-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640135.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640135-supp.pdf +is-appearance-free-action-recognition-possible,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640154.pdf, +learning-spatial-preserved-skeleton-representations-for-few-shot-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640172.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640172-supp.pdf +dual-evidential-learning-for-weakly-supervised-temporal-action-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640190.pdf, +global-local-motion-transformer-for-unsupervised-skeleton-based-action-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640207.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640207-supp.pdf +adafocusv3-on-unified-spatial-temporal-dynamic-video-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640224.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640224-supp.pdf +panoramic-human-activity-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640242.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640242-supp.pdf +delving-into-details-synopsis-to-detail-networks-for-video-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640259.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640259-supp.pdf +a-generalized-robust-framework-for-timestamp-supervision-in-temporal-action-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640276.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640276-supp.pdf +few-shot-action-recognition-with-hierarchical-matching-and-contrastive-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640293.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640293-supp.pdf +privhar-recognizing-human-actions-from-privacy-preserving-lens,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640310.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640310-supp.zip +scale-aware-spatio-temporal-relation-learning-for-video-anomaly-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640328.pdf, +compound-prototype-matching-for-few-shot-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640346.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640346-supp.pdf +continual-3d-convolutional-neural-networks-for-real-time-processing-of-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640364.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640364-supp.pdf +dynamic-spatio-temporal-specialization-learning-for-fine-grained-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640381.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640381-supp.pdf +dynamic-local-aggregation-network-with-adaptive-clusterer-for-anomaly-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640398.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640398-supp.pdf +action-quality-assessment-with-temporal-parsing-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640416.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640416-supp.pdf +entry-flipped-transformer-for-inference-and-prediction-of-participant-behavior,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640433.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640433-supp.zip +pairwise-contrastive-learning-network-for-action-quality-assessment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640450.pdf, +geometric-features-informed-multi-person-human-object-interaction-recognition-in-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640467.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640467-supp.pdf +actionformer-localizing-moments-of-actions-with-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640485.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640485-supp.pdf +socialvae-human-trajectory-prediction-using-timewise-latents,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640504.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640504-supp.pdf +shape-matters-deformable-patch-attack,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640522.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640522-supp.pdf +frequency-domain-model-augmentation-for-adversarial-attack,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640543.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640543-supp.pdf +prior-guided-adversarial-initialization-for-fast-adversarial-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640560.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640560-supp.pdf +enhanced-accuracy-and-robustness-via-multi-teacher-adversarial-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640577.pdf, +lgv-boosting-adversarial-example-transferability-from-large-geometric-vicinity,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640594.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640594-supp.pdf +a-large-scale-multiple-objective-method-for-black-box-attack-against-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640611.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640611-supp.pdf +gradauto-energy-oriented-attack-on-dynamic-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640628.pdf, +a-spectral-view-of-randomized-smoothing-under-common-corruptions-benchmarking-and-improving-certified-robustness,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640645.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640645-supp.pdf +improving-adversarial-robustness-of-3d-point-cloud-classification-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640663.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640663-supp.pdf +learning-extremely-lightweight-and-robust-model-with-differentiable-constraints-on-sparsity-and-condition-number,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640679.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640679-supp.pdf +ribac-towards-robust-and-imperceptible-backdoor-attack-against-compact-dnn,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640697.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640697-supp.pdf +boosting-transferability-of-targeted-adversarial-examples-via-hierarchical-generative-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640714.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136640714-supp.pdf +adaptive-image-transformations-for-transfer-based-adversarial-attack,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650001-supp.pdf +generative-multiplane-images-making-a-2d-gan-3d-aware,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650019-supp.pdf +advdo-realistic-adversarial-attacks-for-trajectory-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650036-supp.pdf +adversarial-contrastive-learning-via-asymmetric-infonce,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650053.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650053-supp.pdf +one-size-does-not-fit-all-data-adaptive-adversarial-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650070-supp.pdf +unicr-universally-approximated-certified-robustness-via-randomized-smoothing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650086.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650086-supp.pdf +hardly-perceptible-trojan-attack-against-neural-networks-with-bit-flips,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650103.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650103-supp.pdf +robust-network-architecture-search-via-feature-distortion-restraining,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650120.pdf, +secretgen-privacy-recovery-on-pre-trained-models-via-distribution-discrimination,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650137.pdf, +triangle-attack-a-query-efficient-decision-based-adversarial-attack,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650153.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650153-supp.pdf +data-free-backdoor-removal-based-on-channel-lipschitzness,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650171.pdf, +black-box-dissector-towards-erasing-based-hard-label-model-stealing-attack,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650188.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650188-supp.pdf +learning-energy-based-models-with-adversarial-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650204.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650204-supp.pdf +adversarial-label-poisoning-attack-on-graph-neural-networks-via-label-propagation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650223.pdf, +revisiting-outer-optimization-in-adversarial-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650240.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650240-supp.pdf +zero-shot-attribute-attacks-on-fine-grained-recognition-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650257.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650257-supp.pdf +towards-effective-and-robust-neural-trojan-defenses-via-input-filtering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650277.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650277-supp.pdf +scaling-adversarial-training-to-large-perturbation-bounds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650295.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650295-supp.pdf +exploiting-the-local-parabolic-landscapes-of-adversarial-losses-to-accelerate-black-box-adversarial-attack,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650311.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650311-supp.pdf +generative-domain-adaptation-for-face-anti-spoofing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650328.pdf, +metagait-learning-to-learn-an-omni-sample-adaptive-representation-for-gait-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650350.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650350-supp.pdf +gaitedge-beyond-plain-end-to-end-gait-recognition-for-better-practicality,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650368.pdf, +uia-vit-unsupervised-inconsistency-aware-method-based-on-vision-transformer-for-face-forgery-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650384.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650384-supp.pdf +effective-presentation-attack-detection-driven-by-face-related-task,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650400.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650400-supp.pdf +ppt-token-pruned-pose-transformer-for-monocular-and-multi-view-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650416.pdf, +avatarposer-articulated-full-body-pose-tracking-from-sparse-motion-sensing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650434.pdf, +p-stmo-pre-trained-spatial-temporal-many-to-one-model-for-3d-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650453.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650453-supp.pdf +d-d-learning-human-dynamics-from-dynamic-camera,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650470.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650470-supp.pdf +explicit-occlusion-reasoning-for-multi-person-3d-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650488.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650488-supp.pdf +couch-towards-controllable-human-chair-interactions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650508.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650508-supp.pdf +identity-aware-hand-mesh-estimation-and-personalization-from-rgb-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650526.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650526-supp.zip +c3p-cross-domain-pose-prior-propagation-for-weakly-supervised-3d-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650544.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650544-supp.pdf +pose-ndf-modeling-human-pose-manifolds-with-neural-distance-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650562.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650562-supp.pdf +cliff-carrying-location-information-in-full-frames-into-human-pose-and-shape-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650580.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650580-supp.pdf +deciwatch-a-simple-baseline-for-10x-efficient-2d-and-3d-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650597.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650597-supp.pdf +smoothnet-a-plug-and-play-network-for-refining-human-poses-in-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650615.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650615-supp.pdf +posetrans-a-simple-yet-effective-pose-transformation-augmentation-for-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650633.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650633-supp.pdf +multi-person-3d-pose-and-shape-estimation-via-inverse-kinematics-and-refinement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650650.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650650-supp.pdf +overlooked-poses-actually-make-sense-distilling-privileged-knowledge-for-human-motion-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650668.pdf, +structural-triangulation-a-closed-form-solution-to-constrained-3d-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650685.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650685-supp.pdf +audio-driven-stylized-gesture-generation-with-flow-based-model,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650701.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650701-supp.zip +self-constrained-inference-optimization-on-structural-groups-for-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650718.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136650718-supp.pdf +unrealego-a-new-dataset-for-robust-egocentric-3d-human-motion-capture,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660001-supp.pdf +skeleton-parted-graph-scattering-networks-for-3d-human-motion-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660018.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660018-supp.pdf +rethinking-keypoint-representations-modeling-keypoints-and-poses-as-objects-for-multi-person-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660036.pdf, +virtualpose-learning-generalizable-3d-human-pose-models-from-virtual-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660054.pdf, +poseur-direct-human-pose-regression-with-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660071.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660071-supp.pdf +simcc-a-simple-coordinate-classification-perspective-for-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660088.pdf, +regularizing-vector-embedding-in-bottom-up-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660105.pdf, +a-visual-navigation-perspective-for-category-level-object-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660121.pdf, +faster-voxelpose-real-time-3d-human-pose-estimation-by-orthographic-projection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660139.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660139-supp.zip +learning-to-fit-morphable-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660156.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660156-supp.pdf +egobody-human-body-shape-and-motion-of-interacting-people-from-head-mounted-devices,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660176.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660176-supp.pdf +graspd-differentiable-contact-rich-grasp-synthesis-for-multi-fingered-hands,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660197.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660197-supp.zip +autoavatar-autoregressive-neural-fields-for-dynamic-avatar-modeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660216.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660216-supp.zip +deep-radial-embedding-for-visual-sequence-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660234.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660234-supp.pdf +saga-stochastic-whole-body-grasping-with-contact,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660251.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660251-supp.pdf +neural-capture-of-animatable-3d-human-from-monocular-video,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660269.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660269-supp.zip +general-object-pose-transformation-network-from-unpaired-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660286.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660286-supp.pdf +compositional-human-scene-interaction-synthesis-with-semantic-control,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660305.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660305-supp.pdf +pressurevision-estimating-hand-pressure-from-a-single-rgb-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660322.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660322-supp.pdf +posescript-3d-human-poses-from-natural-language,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660340.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660340-supp.zip +dprost-dynamic-projective-spatial-transformer-network-for-6d-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660357.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660357-supp.pdf +3d-interacting-hand-pose-estimation-by-hand-de-occlusion-and-removal,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660374.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660374-supp.pdf +pose-for-everything-towards-category-agnostic-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660391.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660391-supp.pdf +posegpt-quantization-based-3d-human-motion-generation-and-forecasting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660409.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660409-supp.zip +dh-aug-dh-forward-kinematics-model-driven-augmentation-for-3d-human-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660427.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660427-supp.pdf +estimating-spatially-varying-lighting-in-urban-scenes-with-disentangled-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660445.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660445-supp.pdf +boosting-event-stream-super-resolution-with-a-recurrent-neural-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660461.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660461-supp.zip +projective-parallel-single-pixel-imaging-to-overcome-global-illumination-in-3d-structure-light-scanning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660479.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660479-supp.pdf +semantic-sparse-colorization-network-for-deep-exemplar-based-colorization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660495.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660495-supp.pdf +practical-and-scalable-desktop-based-high-quality-facial-capture,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660512.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660512-supp.zip +fast-vqa-efficient-end-to-end-video-quality-assessment-with-fragment-sampling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660528.pdf, +physically-based-editing-of-indoor-scene-lighting-from-a-single-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660545.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660545-supp.pdf +lednet-joint-low-light-enhancement-and-deblurring-in-the-dark,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660562.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660562-supp.pdf +mpib-an-mpi-based-bokeh-rendering-framework-for-realistic-partial-occlusion-effects,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660579.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660579-supp.pdf +real-rawvsr-real-world-raw-video-super-resolution-with-a-benchmark-dataset,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660597.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660597-supp.pdf +transform-your-smartphone-into-a-dslr-camera-learning-the-isp-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660614.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660614-supp.pdf +learning-deep-non-blind-image-deconvolution-without-ground-truths,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660631.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660631-supp.pdf +nest-neural-event-stack-for-event-based-image-enhancement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660649.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660649-supp.pdf +editable-indoor-lighting-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660666.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660666-supp.pdf +fast-two-step-blind-optical-aberration-correction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660682.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660682-supp.pdf +seeing-far-in-the-dark-with-patterned-flash,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660698.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660698-supp.pdf +pseudoclick-interactive-image-segmentation-with-click-imitation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660717.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136660717-supp.pdf +ct2-colorization-transformer-via-color-tokens,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670001-supp.pdf +simple-baselines-for-image-restoration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670017.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670017-supp.pdf +spike-transformer-monocular-depth-estimation-for-spiking-camera,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670034.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670034-supp.pdf +improving-image-restoration-by-revisiting-global-information-aggregation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670053.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670053-supp.pdf +data-association-between-event-streams-and-intensity-frames-under-diverse-baselines,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670071.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670071-supp.pdf +d2hnet-joint-denoising-and-deblurring-with-hierarchical-network-for-robust-night-image-restoration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670089.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670089-supp.pdf +learning-graph-neural-networks-for-image-style-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670108.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670108-supp.pdf +deepps2-revisiting-photometric-stereo-using-two-differently-illuminated-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670125.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670125-supp.pdf +instance-contour-adjustment-via-structure-driven-cnn,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670142.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670142-supp.pdf +synthesizing-light-field-video-from-monocular-video,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670158.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670158-supp.zip +human-centric-image-cropping-with-partition-aware-and-content-preserving-features,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670176.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670176-supp.pdf +demfi-deep-joint-deblurring-and-multi-frame-interpolation-with-flow-guided-attentive-correlation-and-recursive-boosting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670193.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670193-supp.pdf +neural-image-representations-for-multi-image-fusion-and-layer-separation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670210.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670210-supp.pdf +bringing-rolling-shutter-images-alive-with-dual-reversed-distortion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670227.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670227-supp.zip +film-frame-interpolation-for-large-motion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670244.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670244-supp.pdf +video-interpolation-by-event-driven-anisotropic-adjustment-of-optical-flow,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670261.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670261-supp.zip +evac3d-from-event-based-apparent-contours-to-3d-models-via-continuous-visual-hulls,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670278.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670278-supp.pdf +dccf-deep-comprehensible-color-filter-learning-framework-for-high-resolution-image-harmonization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670294.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670294-supp.pdf +selectionconv-convolutional-neural-networks-for-non-rectilinear-image-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670310.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670310-supp.pdf +spatial-separated-curve-rendering-network-for-efficient-and-high-resolution-image-harmonization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670327.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670327-supp.pdf +bigcolor-colorization-using-a-generative-color-prior-for-natural-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670343.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670343-supp.pdf +cadyq-content-aware-dynamic-quantization-for-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670360.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670360-supp.pdf +deep-semantic-statistics-matching-d2sm-denoising-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670377.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670377-supp.zip +3d-scene-inference-from-transient-histograms,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670394.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670394-supp.pdf +neural-space-filling-curves,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670412.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670412-supp.pdf +exposure-aware-dynamic-weighted-learning-for-single-shot-hdr-imaging,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670429.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670429-supp.pdf +seeing-through-a-black-box-toward-high-quality-terahertz-imaging-via-subspace-and-attention-guided-restoration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670447.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670447-supp.pdf +tomography-of-turbulence-strength-based-on-scintillation-imaging,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670464.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670464-supp.zip +realistic-blur-synthesis-for-learning-image-deblurring,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670481.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670481-supp.pdf +learning-phase-mask-for-privacy-preserving-passive-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670497.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670497-supp.pdf +lwgnet-learned-wirtinger-gradients-for-fourier-ptychographic-phase-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670515.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670515-supp.pdf +pandora-polarization-aided-neural-decomposition-of-radiance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670531.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670531-supp.zip +humman-multi-modal-4d-human-dataset-for-versatile-sensing-and-modeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670549.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670549-supp.pdf +dvs-voltmeter-stochastic-process-based-event-simulator-for-dynamic-vision-sensors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670571.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670571-supp.pdf +benchmarking-omni-vision-representation-through-the-lens-of-visual-realms,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670587.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670587-supp.zip +beat-a-large-scale-semantic-and-emotional-multi-modal-dataset-for-conversational-gestures-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670605.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670605-supp.pdf +neuromorphic-data-augmentation-for-training-spiking-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670623.pdf, +celebv-hq-a-large-scale-video-facial-attributes-dataset,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670641.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670641-supp.pdf +moviecuts-a-new-dataset-and-benchmark-for-cut-type-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670659.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670659-supp.zip +lamar-benchmarking-localization-and-mapping-for-augmented-reality,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670677.pdf, +unitail-detecting-reading-and-matching-in-retail-scene,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670695.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670695-supp.pdf +not-just-streaks-towards-ground-truth-for-single-image-deraining,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670713.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136670713-supp.pdf +eccv-caption-correcting-false-negatives-by-collecting-machine-and-human-verified-image-caption-associations-for-ms-coco,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680001-supp.pdf +motcom-the-multi-object-tracking-dataset-complexity-metric,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680019-supp.pdf +how-to-synthesize-a-large-scale-and-trainable-micro-expression-dataset,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680037.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680037-supp.pdf +a-real-world-dataset-for-multi-view-3d-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680054.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680054-supp.zip +realy-rethinking-the-evaluation-of-3d-face-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680072-supp.pdf +capturing-reconstructing-and-simulating-the-urbanscene3d-dataset,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680090.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680090-supp.pdf +3d-compat-composition-of-materials-on-parts-of-3d-things,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680107.pdf, +partimagenet-a-large-high-quality-dataset-of-parts,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680124.pdf, +a-okvqa-a-benchmark-for-visual-question-answering-using-world-knowledge,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680141.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680141-supp.pdf +ood-cv-a-benchmark-for-robustness-to-out-of-distribution-shifts-of-individual-nuisances-in-natural-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680158.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680158-supp.pdf +facial-depth-and-normal-estimation-using-single-dual-pixel-camera,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680176.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680176-supp.pdf +the-anatomy-of-video-editing-a-dataset-and-benchmark-suite-for-ai-assisted-video-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680195.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680195-supp.pdf +stylebabel-artistic-style-tagging-and-captioning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680212.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680212-supp.pdf +pandora-a-panoramic-detection-dataset-for-object-with-orientation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680229.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680229-supp.pdf +fs-coco-towards-understanding-of-freehand-sketches-of-common-objects-in-context,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680245.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680245-supp.pdf +exploring-fine-grained-audiovisual-categorization-with-the-ssw60-dataset,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680262.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680262-supp.pdf +the-caltech-fish-counting-dataset-a-benchmark-for-multiple-object-tracking-and-counting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680281.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680281-supp.pdf +a-dataset-for-interactive-vision-language-navigation-with-unknown-command-feasibility,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680304.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680304-supp.pdf +brace-the-breakdancing-competition-dataset-for-dance-motion-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680321.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680321-supp.pdf +dress-code-high-resolution-multi-category-virtual-try-on,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680337.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680337-supp.pdf +a-data-centric-approach-for-improving-ambiguous-labels-with-combined-semi-supervised-classification-and-clustering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680354.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680354-supp.pdf +clearpose-large-scale-transparent-object-dataset-and-benchmark,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680372.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680372-supp.pdf +when-deep-classifiers-agree-analyzing-correlations-between-learning-order-and-image-statistics,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680388.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680388-supp.pdf +animeceleb-large-scale-animation-celebheads-dataset-for-head-reenactment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680405.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680405-supp.pdf +mugen-a-playground-for-video-audio-text-multimodal-understanding-and-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680421.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680421-supp.zip +a-dense-material-segmentation-dataset-for-indoor-and-outdoor-scene-parsing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680440.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680440-supp.pdf +mimicme-a-large-scale-diverse-4d-database-for-facial-expression-analysis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680457.pdf, +delving-into-universal-lesion-segmentation-method-dataset-and-benchmark,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680475.pdf, +large-scale-real-world-multi-person-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680493.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680493-supp.pdf +d2-tpred-discontinuous-dependency-for-trajectory-prediction-under-traffic-lights,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680512.pdf, +the-missing-link-finding-label-relations-across-datasets,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680530.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680530-supp.pdf +learning-omnidirectional-flow-in-360deg-video-via-siamese-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680546.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680546-supp.pdf +vizwiz-fewshot-locating-objects-in-images-taken-by-people-with-visual-impairments,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680563.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680563-supp.pdf +trove-transforming-road-scene-datasets-into-photorealistic-virtual-environments,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680579.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680579-supp.pdf +trapped-in-texture-bias-a-large-scale-comparison-of-deep-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680597.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680597-supp.pdf +deformable-feature-aggregation-for-dynamic-multi-modal-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680616.pdf, +welsa-learning-to-predict-6d-pose-from-weakly-labeled-data-using-shape-alignment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680633.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680633-supp.zip +graph-r-cnn-towards-accurate-3d-object-detection-with-semantic-decorated-local-graph,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680650.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680650-supp.pdf +mppnet-multi-frame-feature-intertwining-with-proxy-points-for-3d-temporal-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680667.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680667-supp.pdf +long-tail-detection-with-effective-class-margins,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680684.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680684-supp.pdf +semi-supervised-monocular-3d-object-detection-by-multi-view-consistency,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680702.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680702-supp.pdf +ptseformer-progressive-temporal-spatial-enhanced-transformer-towards-video-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136680719.pdf, +bevformer-learning-birds-eye-view-representation-from-multi-camera-images-via-spatiotemporal-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690001-supp.pdf +category-level-6d-object-pose-and-size-estimation-using-self-supervised-deep-prior-deformation-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690019.pdf, +dense-teacher-dense-pseudo-labels-for-semi-supervised-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690036.pdf, +point-to-box-network-for-accurate-object-detection-via-single-point-supervision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690053.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690053-supp.pdf +domain-adaptive-hand-keypoint-and-pixel-localization-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690070-supp.pdf +towards-data-efficient-detection-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690090.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690090-supp.pdf +open-vocabulary-detr-with-conditional-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690107.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690107-supp.pdf +prediction-guided-distillation-for-dense-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690123.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690123-supp.pdf +multimodal-object-detection-via-probabilistic-ensembling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690139.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690139-supp.pdf +exploiting-unlabeled-data-with-vision-and-language-models-for-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690156.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690156-supp.pdf +cpo-change-robust-panorama-to-point-cloud-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690173.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690173-supp.pdf +int-towards-infinite-frames-3d-detection-with-an-efficient-framework,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690190.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690190-supp.pdf +end-to-end-weakly-supervised-object-detection-with-sparse-proposal-evolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690207.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690207-supp.pdf +calibration-free-multi-view-crowd-counting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690224.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690224-supp.pdf +unsupervised-domain-adaptation-for-monocular-3d-object-detection-via-self-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690242.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690242-supp.pdf +superline3d-self-supervised-line-segmentation-and-description-for-lidar-point-cloud,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690259.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690259-supp.zip +exploring-plain-vision-transformer-backbones-for-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690276.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690276-supp.pdf +adversarially-aware-robust-object-detector,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690293.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690293-supp.pdf +head-hetero-assists-distillation-for-heterogeneous-object-detectors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690310.pdf, +you-should-look-at-all-objects,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690327.pdf, +detecting-twenty-thousand-classes-using-image-level-supervision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690344.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690344-supp.pdf +dcl-net-deep-correspondence-learning-network-for-6d-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690362.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690362-supp.pdf +monocular-3d-object-detection-with-depth-from-motion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690380.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690380-supp.zip +disp6d-disentangled-implicit-shape-and-pose-learning-for-scalable-6d-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690397.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690397-supp.pdf +distilling-object-detectors-with-global-knowledge,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690415.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690415-supp.pdf +unifying-visual-perception-by-dispersible-points-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690432.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690432-supp.pdf +pseco-pseudo-labeling-and-consistency-training-for-semi-supervised-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690449.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690449-supp.pdf +exploring-resolution-and-degradation-clues-as-self-supervised-signal-for-low-quality-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690465.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690465-supp.pdf +robust-category-level-6d-pose-estimation-with-coarse-to-fine-rendering-of-neural-features,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690484.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690484-supp.pdf +translation-scale-and-rotation-cross-modal-alignment-meets-rgb-infrared-vehicle-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690501.pdf, +rfla-gaussian-receptive-field-based-label-assignment-for-tiny-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690518.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690518-supp.pdf +rethinking-iou-based-optimization-for-single-stage-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690536.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690536-supp.pdf +td-road-top-down-road-network-extraction-with-holistic-graph-construction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690553.pdf, +multi-faceted-distillation-of-base-novel-commonality-for-few-shot-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690569.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690569-supp.pdf +pointclm-a-contrastive-learning-based-framework-for-multi-instance-point-cloud-registration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690586.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690586-supp.pdf +weakly-supervised-object-localization-via-transformer-with-implicit-spatial-calibration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690603.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690603-supp.pdf +mttrans-cross-domain-object-detection-with-mean-teacher-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690620.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690620-supp.pdf +multi-domain-multi-definition-landmark-localization-for-small-datasets,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690637.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690637-supp.pdf +deviant-depth-equivariant-network-for-monocular-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690655.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690655-supp.pdf +label-guided-auxiliary-training-improves-3d-object-detector,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690674.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690674-supp.pdf +promptdet-towards-open-vocabulary-detection-using-uncurated-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690691.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690691-supp.pdf +densely-constrained-depth-estimator-for-monocular-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690708.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690708-supp.pdf +polarimetric-pose-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690726.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136690726-supp.pdf +dfnet-enhance-absolute-pose-regression-with-direct-feature-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700001-supp.pdf +cornerformer-purifying-instances-for-corner-based-detectors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700017.pdf, +pillarnet-real-time-and-high-performance-pillar-based-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700034.pdf, +robust-object-detection-with-inaccurate-bounding-boxes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700052.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700052-supp.pdf +efficient-decoder-free-object-detection-with-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700069.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700069-supp.pdf +cross-modality-knowledge-distillation-network-for-monocular-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700085.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700085-supp.pdf +react-temporal-action-detection-with-relational-queries,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700102.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700102-supp.pdf +towards-accurate-active-camera-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700119.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700119-supp.pdf +camera-pose-auto-encoders-for-improving-pose-regression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700137.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700137-supp.pdf +improving-the-intra-class-long-tail-in-3d-detection-via-rare-example-mining,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700155.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700155-supp.pdf +bagging-regional-classification-activation-maps-for-weakly-supervised-object-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700174.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700174-supp.zip +uc-owod-unknown-classified-open-world-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700191.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700191-supp.pdf +raytran-3d-pose-estimation-and-shape-reconstruction-of-multiple-objects-from-videos-with-ray-traced-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700209.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700209-supp.pdf +gtcar-graph-transformer-for-camera-re-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700227.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700227-supp.pdf +3d-object-detection-with-a-self-supervised-lidar-scene-flow-backbone,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700244.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700244-supp.pdf +open-vocabulary-object-detection-with-pseudo-bounding-box-labels,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700263.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700263-supp.pdf +few-shot-object-detection-by-knowledge-distillation-using-bag-of-visual-words-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700279.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700279-supp.pdf +salisa-saliency-based-input-sampling-for-efficient-video-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700296.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700296-supp.pdf +eco-tr-efficient-correspondences-finding-via-coarse-to-fine-refinement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700313.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700313-supp.pdf +vote-from-the-center-6-dof-pose-estimation-in-rgb-d-images-by-radial-keypoint-voting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700331.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700331-supp.pdf +long-tailed-instance-segmentation-using-gumbel-optimized-loss,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700349.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700349-supp.pdf +detmatch-two-teachers-are-better-than-one-for-joint-2d-and-3d-semi-supervised-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700366.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700366-supp.pdf +objectbox-from-centers-to-boxes-for-anchor-free-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700385.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700385-supp.pdf +is-geometry-enough-for-matching-in-visual-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700402.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700402-supp.pdf +swformer-sparse-window-transformer-for-3d-object-detection-in-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700422.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700422-supp.pdf +pcr-cg-point-cloud-registration-via-deep-explicit-color-and-geometry,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700439.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700439-supp.pdf +glamd-global-and-local-attention-mask-distillation-for-object-detectors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700456.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700456-supp.zip +fcaf3d-fully-convolutional-anchor-free-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700473.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700473-supp.pdf +video-anomaly-detection-by-solving-decoupled-spatio-temporal-jigsaw-puzzles,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700490.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700490-supp.pdf +class-agnostic-object-detection-with-multi-modal-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700507.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700507-supp.pdf +enhancing-multi-modal-features-using-local-self-attention-for-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700527.pdf, +object-detection-as-probabilistic-set-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700545.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700545-supp.pdf +weakly-supervised-temporal-action-detection-for-fine-grained-videos-with-hierarchical-atomic-actions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700562.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700562-supp.pdf +neural-correspondence-field-for-object-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700580.pdf, +on-label-granularity-and-object-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700598.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700598-supp.pdf +oimnet-prototypical-normalization-and-localization-aware-learning-for-person-search,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700615.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700615-supp.pdf +out-of-distribution-identification-let-detector-tell-which-i-am-not-sure,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700631.pdf, +learning-with-free-object-segments-for-long-tailed-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700648.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700648-supp.pdf +autoregressive-uncertainty-modeling-for-3d-bounding-box-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700665.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700665-supp.pdf +3d-random-occlusion-and-multi-layer-projection-for-deep-multi-camera-pedestrian-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700681.pdf, +a-simple-single-scale-vision-transformer-for-object-detection-and-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700697.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700697-supp.pdf +simple-open-vocabulary-object-detection-with-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700714.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136700714-supp.pdf +a-simple-approach-and-benchmark-for-21000-category-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710001.pdf, +knowledge-condensation-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710019-supp.pdf +reducing-information-loss-for-spiking-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710036.pdf, +masked-generative-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710053.pdf, +fine-grained-data-distribution-alignment-for-post-training-quantization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710070-supp.pdf +learning-with-recoverable-forgetting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710087.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710087-supp.zip +efficient-one-pass-self-distillation-with-zipfs-label-smoothing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710104.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710104-supp.pdf +prune-your-model-before-distill-it,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710120.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710120-supp.pdf +deep-partial-updating-towards-communication-efficient-updating-for-on-device-inference,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710137.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710137-supp.pdf +patch-similarity-aware-data-free-quantization-for-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710154.pdf, +l3-accelerator-friendly-lossless-image-format-for-high-resolution-high-throughput-dnn-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710171.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710171-supp.pdf +streaming-multiscale-deep-equilibrium-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710189.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710189-supp.pdf +symmetry-regularization-and-saturating-nonlinearity-for-robust-quantization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710207.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710207-supp.pdf +sp-net-slowly-progressing-dynamic-inference-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710225.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710225-supp.pdf +equivariance-and-invariance-inductive-bias-for-learning-from-insufficient-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710242.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710242-supp.pdf +mixed-precision-neural-network-quantization-via-learned-layer-wise-importance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710260.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710260-supp.pdf +event-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710276.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710276-supp.zip +edgevits-competing-light-weight-cnns-on-mobile-devices-with-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710294.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710294-supp.pdf +palquant-accelerating-high-precision-networks-on-low-precision-accelerators,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710312.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710312-supp.pdf +disentangled-differentiable-network-pruning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710329.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710329-supp.pdf +ida-det-an-information-discrepancy-aware-distillation-for-1-bit-detectors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710347.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710347-supp.pdf +learning-to-weight-samples-for-dynamic-early-exiting-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710363.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710363-supp.pdf +adabin-improving-binary-neural-networks-with-adaptive-binary-sets,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710380.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710380-supp.pdf +adaptive-token-sampling-for-efficient-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710397.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710397-supp.pdf +weight-fixing-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710416.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710416-supp.pdf +self-slimmed-vision-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710433.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710433-supp.pdf +switchable-online-knowledge-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710450.pdf, +l-robustness-and-beyond-unleashing-efficient-adversarial-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710466.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710466-supp.pdf +multi-granularity-pruning-for-model-acceleration-on-mobile-devices,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710483.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710483-supp.pdf +deep-ensemble-learning-by-diverse-knowledge-distillation-for-fine-grained-object-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710501.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710501-supp.pdf +helpful-or-harmful-inter-task-association-in-continual-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710518.pdf, +towards-accurate-binary-neural-networks-via-modeling-contextual-dependencies,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710535.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710535-supp.pdf +spin-an-empirical-evaluation-on-sharing-parameters-of-isotropic-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710552.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710552-supp.pdf +ensemble-knowledge-guided-sub-network-search-and-fine-tuning-for-filter-pruning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710568.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710568-supp.pdf +network-binarization-via-contrastive-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710585.pdf, +lipschitz-continuity-retained-binary-neural-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710601.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710601-supp.pdf +spvit-enabling-faster-vision-transformers-via-latency-aware-soft-token-pruning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710618.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710618-supp.pdf +soft-masking-for-cost-constrained-channel-pruning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710640.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710640-supp.pdf +non-uniform-step-size-quantization-for-accurate-post-training-quantization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710657.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710657-supp.pdf +supertickets-drawing-task-agnostic-lottery-tickets-from-supernets-via-jointly-architecture-searching-and-parameter-pruning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710673.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710673-supp.pdf +meta-gf-training-dynamic-depth-neural-networks-harmoniously,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710691.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710691-supp.pdf +towards-ultra-low-latency-spiking-neural-networks-for-vision-and-sequential-tasks-using-temporal-pruning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710709.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710709-supp.zip +towards-accurate-network-quantization-with-equivalent-smooth-regularizer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710726.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136710726-supp.pdf +explicit-model-size-control-and-relaxation-via-smooth-regularization-for-mixed-precision-quantization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720001-supp.pdf +basq-branch-wise-activation-clipping-search-quantization-for-sub-4-bit-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720017.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720017-supp.pdf +you-already-have-it-a-generator-free-low-precision-dnn-training-framework-using-stochastic-rounding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720034.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720034-supp.pdf +real-spike-learning-real-valued-spikes-for-spiking-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720052.pdf, +fedltn-federated-learning-for-sparse-and-personalized-lottery-ticket-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720069.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720069-supp.pdf +theoretical-understanding-of-the-information-flow-on-continual-learning-performance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720085.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720085-supp.pdf +exploring-lottery-ticket-hypothesis-in-spiking-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720101.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720101-supp.pdf +on-the-angular-update-and-hyperparameter-tuning-of-a-scale-invariant-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720120.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720120-supp.pdf +lana-latency-aware-network-acceleration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720136.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720136-supp.pdf +rdo-q-extremely-fine-grained-channel-wise-quantization-via-rate-distortion-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720156.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720156-supp.pdf +u-boost-nas-utilization-boosted-differentiable-neural-architecture-search,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720172.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720172-supp.pdf +ptq4vit-post-training-quantization-for-vision-transformers-with-twin-uniform-quantization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720190.pdf, +bitwidth-adaptive-quantization-aware-neural-network-training-a-meta-learning-approach,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720207.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720207-supp.pdf +understanding-the-dynamics-of-dnns-using-graph-modularity,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720224.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720224-supp.pdf +latent-discriminant-deterministic-uncertainty,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720242.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720242-supp.pdf +making-heads-or-tails-towards-semantically-consistent-visual-counterfactuals,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720260.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720260-supp.pdf +hive-evaluating-the-human-interpretability-of-visual-explanations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720277.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720277-supp.pdf +bayescap-bayesian-identity-cap-for-calibrated-uncertainty-in-frozen-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720295.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720295-supp.pdf +sess-saliency-enhancing-with-scaling-and-sliding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720313.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720313-supp.pdf +no-token-left-behind-explainability-aided-image-classification-and-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720329.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720329-supp.pdf +interpretable-image-classification-with-differentiable-prototypes-assignment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720346.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720346-supp.zip +contributions-of-shape-texture-and-color-in-visual-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720364.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720364-supp.pdf +steex-steering-counterfactual-explanations-with-semantics,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720382.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720382-supp.pdf +are-vision-transformers-robust-to-patch-perturbations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720399.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720399-supp.pdf +a-dataset-generation-framework-for-evaluating-megapixel-image-classifiers-their-explanations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720416.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720416-supp.pdf +cartoon-explanations-of-image-classifiers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720439.pdf, +shap-cam-visual-explanations-for-convolutional-neural-networks-based-on-shapley-value,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720455.pdf, +privacy-preserving-face-recognition-with-learnable-privacy-budgets-in-frequency-domain,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720471.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720471-supp.pdf +contrast-phys-unsupervised-video-based-remote-physiological-measurement-via-spatiotemporal-contrast,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720488.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720488-supp.pdf +source-free-domain-adaptation-with-contrastive-domain-alignment-and-self-supervised-exploration-for-face-anti-spoofing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720506.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720506-supp.pdf +on-mitigating-hard-clusters-for-face-clustering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720523.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720523-supp.pdf +oneface-one-threshold-for-all,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720539.pdf, +label2label-a-language-modeling-framework-for-multi-attribute-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720556.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720556-supp.pdf +agetransgan-for-facial-age-transformation-with-rectified-performance-metrics,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720573.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720573-supp.pdf +hierarchical-contrastive-inconsistency-learning-for-deepfake-video-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720588.pdf, +rethinking-robust-representation-learning-under-fine-grained-noisy-faces,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720605.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720605-supp.pdf +teaching-where-to-look-attention-similarity-knowledge-distillation-for-low-resolution-face-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720622.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720622-supp.pdf +teaching-with-soft-label-smoothing-for-mitigating-noisy-labels-in-facial-expressions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720639.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720639-supp.pdf +learning-dynamic-facial-radiance-fields-for-few-shot-talking-head-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720657.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720657-supp.zip +coupleface-relation-matters-for-face-recognition-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720674.pdf, +controllable-and-guided-face-synthesis-for-unconstrained-face-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720692.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720692-supp.pdf +towards-robust-face-recognition-with-comprehensive-search,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720711.pdf, +towards-unbiased-label-distribution-learning-for-facial-pose-estimation-using-anisotropic-spherical-gaussian,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136720728.pdf, +au-aware-3d-face-reconstruction-through-personalized-au-specific-blendshape-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730001-supp.pdf +bezierpalm-a-free-lunch-for-palmprint-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730019-supp.pdf +adaptive-transformers-for-robust-few-shot-cross-domain-face-anti-spoofing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730037.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730037-supp.pdf +face2facer-real-time-high-resolution-one-shot-face-reenactment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730055.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730055-supp.zip +towards-racially-unbiased-skin-tone-estimation-via-scene-disambiguation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730072-supp.pdf +boundaryface-a-mining-framework-with-noise-label-self-correction-for-face-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730092.pdf, +pre-training-strategies-and-datasets-for-facial-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730109.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730109-supp.pdf +look-both-ways-self-supervising-driver-gaze-estimation-and-road-scene-saliency,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730128.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730128-supp.pdf +mfim-megapixel-facial-identity-manipulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730145.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730145-supp.pdf +3d-face-reconstruction-with-dense-landmarks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730162.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730162-supp.pdf +emotion-aware-multi-view-contrastive-learning-for-facial-emotion-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730181.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730181-supp.zip +order-learning-using-partially-ordered-data-via-chainization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730199.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730199-supp.pdf +unsupervised-high-fidelity-facial-texture-generation-and-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730215.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730215-supp.pdf +multi-domain-learning-for-updating-face-anti-spoofing-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730232.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730232-supp.zip +towards-metrical-reconstruction-of-human-faces,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730249.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730249-supp.zip +discover-and-mitigate-unknown-biases-with-debiasing-alternate-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730270.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730270-supp.pdf +unsupervised-and-semi-supervised-bias-benchmarking-in-face-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730288.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730288-supp.pdf +towards-efficient-adversarial-training-on-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730307.pdf, +mime-minority-inclusion-for-majority-group-enhancement-of-ai-performance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730327.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730327-supp.pdf +studying-bias-in-gans-through-the-lens-of-race,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730345.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730345-supp.pdf +trust-but-verify-using-self-supervised-probing-to-improve-trustworthiness,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730362.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730362-supp.pdf +learning-to-censor-by-noisy-sampling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730378.pdf, +an-invisible-black-box-backdoor-attack-through-frequency-domain,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730396.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730396-supp.pdf +fairgrape-fairness-aware-gradient-pruning-method-for-face-attribute-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730414.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730414-supp.pdf +attaining-class-level-forgetting-in-pretrained-model-using-few-samples,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730433.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730433-supp.zip +anti-neuron-watermarking-protecting-personal-data-against-unauthorized-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730449.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730449-supp.zip +an-impartial-take-to-the-cnn-vs-transformer-robustness-contest,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730466.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730466-supp.pdf +recover-fair-deep-classification-models-via-altering-pre-trained-structure,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730482.pdf, +decouple-and-sample-protecting-sensitive-information-in-task-agnostic-data-release,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730499.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730499-supp.pdf +privacy-preserving-action-recognition-via-motion-difference-quantization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730518.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730518-supp.pdf +latent-space-smoothing-for-individually-fair-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730535.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730535-supp.pdf +parameterized-temperature-scaling-for-boosting-the-expressive-power-in-post-hoc-uncertainty-calibration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730554.pdf, +fairstyle-debiasing-stylegan2-with-style-channel-manipulations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730569.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730569-supp.pdf +distilling-the-undistillable-learning-from-a-nasty-teacher,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730586.pdf, +sos-self-supervised-learning-over-sets-of-handled-objects-in-egocentric-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730603.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730603-supp.pdf +egocentric-activity-recognition-and-localization-on-a-3d-map,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730620.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730620-supp.pdf +generative-adversarial-network-for-future-hand-segmentation-from-egocentric-video,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730638.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730638-supp.zip +my-view-is-the-best-view-procedure-learning-from-egocentric-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730656.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730656-supp.pdf +gimo-gaze-informed-human-motion-prediction-in-context,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730675.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730675-supp.pdf +image-based-clip-guided-essence-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730693.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730693-supp.pdf +detecting-and-recovering-sequential-deepfake-manipulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730710.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730710-supp.pdf +self-supervised-sparse-representation-for-video-anomaly-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730727.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136730727-supp.pdf +watermark-vaccine-adversarial-attacks-to-prevent-watermark-removal,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740001-supp.pdf +explaining-deepfake-detection-by-analysing-image-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740018.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740018-supp.pdf +frequencylowcut-pooling-plug-play-against-catastrophic-overfitting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740036-supp.pdf +tafim-targeted-adversarial-attacks-against-facial-image-manipulations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740053.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740053-supp.pdf +fingerprintnet-synthesized-fingerprints-for-generated-image-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740071.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740071-supp.pdf +detecting-generated-images-by-real-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740089.pdf, +an-information-theoretic-approach-for-attention-driven-face-forgery-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740105.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740105-supp.pdf +exploring-disentangled-content-information-for-face-forgery-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740122.pdf, +repmix-representation-mixing-for-robust-attribution-of-synthesized-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740140.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740140-supp.pdf +totems-physical-objects-for-verifying-visual-integrity,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740158.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740158-supp.pdf +dual-stream-knowledge-preserving-hashing-for-unsupervised-video-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740175.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740175-supp.pdf +pass-part-aware-self-supervised-pre-training-for-person-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740192.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740192-supp.zip +adaptive-cross-domain-learning-for-generalizable-person-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740209.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740209-supp.pdf +multi-query-video-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740227.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740227-supp.zip +hierarchical-average-precision-training-for-pertinent-image-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740244.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740244-supp.pdf +learning-semantic-correspondence-with-sparse-annotations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740261.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740261-supp.pdf +dynamically-transformed-instance-normalization-network-for-generalizable-person-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740279.pdf, +domain-adaptive-person-search,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740295.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740295-supp.pdf +ts2-net-token-shift-and-selection-transformer-for-text-video-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740311.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740311-supp.pdf +unstructured-feature-decoupling-for-vehicle-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740328.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740328-supp.pdf +deep-hash-distillation-for-image-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740345.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740345-supp.pdf +mimic-embedding-via-adaptive-aggregation-learning-generalizable-person-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740362.pdf, +granularity-aware-adaptation-for-image-retrieval-over-multiple-tasks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740379.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740379-supp.pdf +learning-audio-video-modalities-from-image-captions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740396.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740396-supp.pdf +rvsl-robust-vehicle-similarity-learning-in-real-hazy-scenes-based-on-semi-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740415.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740415-supp.pdf +lightweight-attentional-feature-fusion-a-new-baseline-for-text-to-video-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740432.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740432-supp.pdf +modality-synergy-complement-learning-with-cascaded-aggregation-for-visible-infrared-person-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740450.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740450-supp.pdf +cross-modality-transformer-for-visible-infrared-person-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740467.pdf, +audio-visual-mismatch-aware-video-retrieval-via-association-and-adjustment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740484.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740484-supp.pdf +connecting-compression-spaces-with-transformer-for-approximate-nearest-neighbor-search,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740502.pdf, +semicon-a-learning-to-hash-solution-for-large-scale-fine-grained-image-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740518.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740518-supp.pdf +cavit-contextual-alignment-vision-transformer-for-video-object-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740535.pdf, +text-based-temporal-localization-of-novel-events,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740552.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740552-supp.pdf +reliability-aware-prediction-via-uncertainty-learning-for-person-image-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740572.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740572-supp.pdf +relighting4d-neural-relightable-human-from-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740589.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740589-supp.pdf +real-time-intermediate-flow-estimation-for-video-frame-interpolation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740608.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740608-supp.pdf +pixelfolder-an-efficient-progressive-pixel-synthesis-network-for-image-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740626.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740626-supp.pdf +styleswap-style-based-generator-empowers-robust-face-swapping,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740644.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740644-supp.zip +paint2pix-interactive-painting-based-progressive-image-synthesis-and-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740662.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740662-supp.pdf +furrygan-high-quality-foreground-aware-image-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740679.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740679-supp.pdf +scam-transferring-humans-between-images-with-semantic-cross-attention-modulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740696.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740696-supp.pdf +sem2nerf-converting-single-view-semantic-masks-to-neural-radiance-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740713.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136740713-supp.pdf +wavegan-frequency-aware-gan-for-high-fidelity-few-shot-image-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750001-supp.pdf +end-to-end-visual-editing-with-a-generatively-pre-trained-artist,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750018.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750018-supp.pdf +high-fidelity-gan-inversion-with-padding-space,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750036-supp.pdf +designing-one-unified-framework-for-high-fidelity-face-reenactment-and-swapping,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750053.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750053-supp.pdf +sobolev-training-for-implicit-neural-representations-with-approximated-image-derivatives,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750070-supp.pdf +make-a-scene-scene-based-text-to-image-generation-with-human-priors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750087.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750087-supp.pdf +3d-fm-gan-towards-3d-controllable-face-manipulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750106.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750106-supp.pdf +multi-curve-translator-for-high-resolution-photorealistic-image-translation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750124.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750124-supp.pdf +deep-bayesian-video-frame-interpolation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750141.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750141-supp.pdf +cross-attention-based-style-distribution-for-controllable-person-image-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750158.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750158-supp.zip +keypointnerf-generalizing-image-based-volumetric-avatars-using-relative-spatial-encoding-of-keypoints,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750176.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750176-supp.pdf +viewformer-nerf-free-neural-rendering-from-few-images-using-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750195.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750195-supp.pdf +l-tracing-fast-light-visibility-estimation-on-neural-surfaces-by-sphere-tracing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750214.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750214-supp.pdf +a-perceptual-quality-metric-for-video-frame-interpolation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750231.pdf, +adaptive-feature-interpolation-for-low-shot-image-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750251.pdf, +palgan-image-colorization-with-palette-generative-adversarial-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750268.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750268-supp.pdf +fast-vid2vid-spatial-temporal-compression-for-video-to-video-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750285.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750285-supp.pdf +learning-prior-feature-and-attention-enhanced-image-inpainting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750303.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750303-supp.pdf +temporal-mpi-enabling-multi-plane-images-for-dynamic-scene-modelling-via-temporal-basis-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750321.pdf, +3d-aware-semantic-guided-generative-model-for-human-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750337.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750337-supp.pdf +temporally-consistent-semantic-video-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750355.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750355-supp.pdf +error-compensation-framework-for-flow-guided-video-inpainting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750373.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750373-supp.pdf +scraping-textures-from-natural-images-for-synthesis-and-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750389.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750389-supp.pdf +single-stage-virtual-try-on-via-deformable-attention-flows,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750406.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750406-supp.pdf +improving-gans-for-long-tailed-data-through-group-spectral-regularization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750423.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750423-supp.pdf +hierarchical-semantic-regularization-of-latent-spaces-in-stylegans,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750440.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750440-supp.pdf +interestyle-encoding-an-interest-region-for-robust-stylegan-inversion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750457.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750457-supp.pdf +stylelight-hdr-panorama-generation-for-lighting-estimation-and-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750474.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750474-supp.pdf +contrastive-monotonic-pixel-level-modulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750491.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750491-supp.pdf +learning-cross-video-neural-representations-for-high-quality-frame-interpolation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750509.pdf, +learning-continuous-implicit-representation-for-near-periodic-patterns,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750527.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750527-supp.pdf +end-to-end-graph-constrained-vectorized-floorplan-generation-with-panoptic-refinement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750545.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750545-supp.pdf +few-shot-image-generation-with-mixup-based-distance-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750561.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750561-supp.pdf +a-style-based-gan-encoder-for-high-fidelity-reconstruction-of-images-and-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750579.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750579-supp.pdf +fakeclr-exploring-contrastive-learning-for-solving-latent-discontinuity-in-data-efficient-gans,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750596.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750596-supp.pdf +blobgan-spatially-disentangled-scene-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750613.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750613-supp.pdf +unified-implicit-neural-stylization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750633.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750633-supp.pdf +gan-with-multivariate-disentangling-for-controllable-hair-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750653.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750653-supp.pdf +discovering-transferable-forensic-features-for-cnn-generated-images-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750669.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750669-supp.pdf +harmonizer-learning-to-perform-white-box-image-and-video-harmonization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750688.pdf, +text2live-text-driven-layered-image-and-video-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750705.pdf, +digging-into-radiance-grid-for-real-time-view-synthesis-with-detail-preservation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136750722.pdf, +stylegan-human-a-data-centric-odyssey-of-human-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760001-supp.pdf +colorformer-image-colorization-via-color-memory-assisted-hybrid-attention-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760020.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760020-supp.pdf +eagan-efficient-two-stage-evolutionary-architecture-search-for-gans,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760036.pdf, +weakly-supervised-stitching-network-for-real-world-panoramic-image-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760052.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760052-supp.pdf +dynast-dynamic-sparse-transformer-for-exemplar-guided-image-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760070-supp.pdf +multimodal-conditional-image-synthesis-with-product-of-experts-gans,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760089.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760089-supp.pdf +auto-regressive-image-synthesis-with-integrated-quantization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760106.pdf, +jojogan-one-shot-face-stylization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760124.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760124-supp.pdf +vecgan-image-to-image-translation-with-interpretable-latent-directions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760141.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760141-supp.pdf +any-resolution-training-for-high-resolution-image-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760158.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760158-supp.pdf +ccpl-contrastive-coherence-preserving-loss-for-versatile-style-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760176.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760176-supp.pdf +canf-vc-conditional-augmented-normalizing-flows-for-video-compression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760193.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760193-supp.pdf +bi-level-feature-alignment-for-versatile-image-translation-and-manipulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760210.pdf, +high-fidelity-image-inpainting-with-gan-inversion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760228.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760228-supp.pdf +deltagan-towards-diverse-few-shot-image-generation-with-sample-specific-delta,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760245.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760245-supp.pdf +image-inpainting-with-cascaded-modulation-gan-and-object-aware-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760263.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760263-supp.pdf +styleface-towards-identity-disentangled-face-generation-on-megapixels,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760281.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760281-supp.pdf +video-extrapolation-in-space-and-time,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760297.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760297-supp.pdf +contrastive-learning-for-diverse-disentangled-foreground-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760313.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760313-supp.pdf +bips-bi-modal-indoor-panorama-synthesis-via-residual-depth-aided-adversarial-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760331.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760331-supp.pdf +augmentation-of-rppg-benchmark-datasets-learning-to-remove-and-embed-rppg-signals-via-double-cycle-consistent-learning-from-unpaired-facial-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760351.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760351-supp.zip +geometry-aware-single-image-full-body-human-relighting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760367.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760367-supp.pdf +3d-aware-indoor-scene-synthesis-with-depth-priors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760385.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760385-supp.pdf +deep-portrait-delighting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760402.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760402-supp.zip +vector-quantized-image-to-image-translation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760419.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760419-supp.pdf +the-surprisingly-straightforward-scene-text-removal-method-with-gated-attention-and-region-of-interest-generation-a-comprehensive-prominent-model-analysis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760436.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760436-supp.pdf +free-viewpoint-rgb-d-human-performance-capture-and-rendering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760452.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760452-supp.pdf +multiview-regenerative-morphing-with-dual-flows,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760469.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760469-supp.pdf +hallucinating-pose-compatible-scenes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760487.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760487-supp.pdf +motion-and-appearance-adaptation-for-cross-domain-motion-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760506.pdf, +layered-controllable-video-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760523.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760523-supp.pdf +custom-structure-preservation-in-face-aging,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760541.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760541-supp.pdf +spatio-temporal-deformable-attention-network-for-video-deblurring,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760558.pdf, +neumesh-learning-disentangled-neural-mesh-based-implicit-field-for-geometry-and-texture-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760574.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760574-supp.zip +nerf-for-outdoor-scene-relighting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760593.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760593-supp.zip +cogs-controllable-generation-and-search-from-sketch-and-style,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760610.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760610-supp.pdf +hairnet-hairstyle-transfer-with-pose-changes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760628.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760628-supp.pdf +unbiased-multi-modality-guidance-for-image-inpainting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760645.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760645-supp.pdf +intelli-paint-towards-developing-more-human-intelligible-painting-agents,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760662.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760662-supp.pdf +motion-transformer-for-unsupervised-image-animation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760679.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760679-supp.pdf +nuwa-visual-synthesis-pre-training-for-neural-visual-world-creation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760697.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760697-supp.pdf +elegant-exquisite-and-locally-editable-gan-for-makeup-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760714.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136760714-supp.pdf +editing-out-of-domain-gan-inversion-via-differential-activations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770001-supp.zip +on-the-robustness-of-quality-measures-for-gans,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770018.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770018-supp.pdf +sound-guided-semantic-video-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770034.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770034-supp.pdf +inpainting-at-modern-camera-resolution-by-guided-patchmatch-with-auto-curation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770051.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770051-supp.pdf +controllable-video-generation-through-global-and-local-motion-dynamics,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770069.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770069-supp.pdf +styleheat-one-shot-high-resolution-editable-talking-face-generation-via-pre-trained-stylegan,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770086.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770086-supp.pdf +long-video-generation-with-time-agnostic-vqgan-and-time-sensitive-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770103.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770103-supp.pdf +combining-internal-and-external-constraints-for-unrolling-shutter-in-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770120.pdf, +wise-whitebox-image-stylization-by-example-based-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770136.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770136-supp.pdf +neural-radiance-transfer-fields-for-relightable-novel-view-synthesis-with-global-illumination,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770155.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770155-supp.zip +transformers-as-meta-learners-for-implicit-neural-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770173.pdf, +style-your-hair-latent-optimization-for-pose-invariant-hairstyle-transfer-via-local-style-aware-hair-alignment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770191.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770191-supp.pdf +high-resolution-virtual-try-on-with-misalignment-and-occlusion-handled-conditions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770208.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770208-supp.pdf +a-codec-information-assisted-framework-for-efficient-compressed-video-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770224.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770224-supp.pdf +injecting-3d-perception-of-controllable-nerf-gan-into-stylegan-for-editable-portrait-image-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770240.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770240-supp.pdf +adanerf-adaptive-sampling-for-real-time-rendering-of-neural-radiance-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770258.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770258-supp.pdf +improving-the-perceptual-quality-of-2d-animation-interpolation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770275.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770275-supp.zip +selective-transhdr-transformer-based-selective-hdr-imaging-using-ghost-region-mask,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770292.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770292-supp.pdf +learning-series-parallel-lookup-tables-for-efficient-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770309.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770309-supp.pdf +geoaug-data-augmentation-for-few-shot-nerf-with-geometry-constraints,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770326.pdf, +doodleformer-creative-sketch-drawing-with-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770343.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770343-supp.pdf +implicit-neural-representations-for-variable-length-human-motion-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770359.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770359-supp.pdf +learning-object-placement-via-dual-path-graph-completion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770376.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770376-supp.pdf +expanded-adaptive-scaling-normalization-for-end-to-end-image-compression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770392.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770392-supp.pdf +generator-knows-what-discriminator-should-learn-in-unconditional-gans,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770408.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770408-supp.pdf +compositional-visual-generation-with-composable-diffusion-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770426.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770426-supp.pdf +manifest-manifold-deformation-for-few-shot-image-translation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770443.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770443-supp.zip +supervised-attribute-information-removal-and-reconstruction-for-image-manipulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770460.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770460-supp.pdf +blt-bidirectional-layout-transformer-for-controllable-layout-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770477.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770477-supp.pdf +diverse-generation-from-a-single-video-made-possible,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770494.pdf, +rayleigh-eigendirections-reds-nonlinear-gan-latent-space-traversals-for-multidimensional-features,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770513.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770513-supp.pdf +bridging-the-domain-gap-towards-generalization-in-automatic-colorization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770530.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770530-supp.pdf +generating-natural-images-with-direct-patch-distributions-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770547.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770547-supp.pdf +context-consistent-semantic-image-editing-with-style-preserved-modulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770564.pdf, +eliminating-gradient-conflict-in-reference-based-line-art-colorization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770582.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770582-supp.pdf +unsupervised-learning-of-efficient-geometry-aware-neural-articulated-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770600.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770600-supp.pdf +jpeg-artifacts-removal-via-contrastive-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770618.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770618-supp.pdf +unpaired-deep-image-dehazing-using-contrastive-disentanglement-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770636.pdf, +efficient-long-range-attention-network-for-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770653.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770653-supp.pdf +flowformer-a-transformer-architecture-for-optical-flow,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770672.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770672-supp.zip +coarse-to-fine-sparse-transformer-for-hyperspectral-image-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770690.pdf, +learning-shadow-correspondence-for-video-shadow-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770709.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770709-supp.pdf +metric-learning-based-interactive-modulation-for-real-world-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770727.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136770727-supp.pdf +dynamic-dual-trainable-bounds-for-ultra-low-precision-super-resolution-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780001-supp.pdf +osformer-one-stage-camouflaged-instance-segmentation-with-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780019.pdf, +highly-accurate-dichotomous-image-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780036-supp.pdf +boosting-supervised-dehazing-methods-via-bi-level-patch-reweighting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780055.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780055-supp.pdf +flow-guided-transformer-for-video-inpainting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780072-supp.pdf +shift-tolerant-perceptual-similarity-metric,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780089.pdf, +perception-distortion-balanced-admm-optimization-for-single-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780106.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780106-supp.pdf +vqfr-blind-face-restoration-with-vector-quantized-dictionary-and-parallel-decoder,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780124.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780124-supp.pdf +uncertainty-learning-in-kernel-estimation-for-multi-stage-blind-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780141.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780141-supp.pdf +learning-spatio-temporal-downsampling-for-effective-video-upscaling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780159.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780159-supp.pdf +learning-local-implicit-fourier-representation-for-image-warping,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780179.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780179-supp.pdf +seplut-separable-image-adaptive-lookup-tables-for-real-time-image-enhancement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780197.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780197-supp.pdf +blind-image-decomposition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780214.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780214-supp.pdf +mulut-cooperating-multiple-look-up-tables-for-efficient-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780234.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780234-supp.pdf +learning-spatiotemporal-frequency-transformer-for-compressed-video-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780252.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780252-supp.pdf +spatial-frequency-domain-information-integration-for-pan-sharpening,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780268.pdf, +adaptive-patch-exiting-for-scalable-single-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780286.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780286-supp.pdf +efficient-meta-tuning-for-content-aware-neural-video-delivery,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780302.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780302-supp.pdf +reference-based-image-super-resolution-with-deformable-attention-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780318.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780318-supp.pdf +local-color-distributions-prior-for-image-enhancement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780336.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780336-supp.pdf +l-coder-language-based-colorization-with-color-object-decoupling-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780352.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780352-supp.pdf +from-face-to-natural-image-learning-real-degradation-for-blind-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780368.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780368-supp.pdf +towards-interpretable-video-super-resolution-via-alternating-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780385.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780385-supp.pdf +event-based-fusion-for-motion-deblurring-with-cross-modal-attention,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780403.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780403-supp.pdf +fast-and-high-quality-image-denoising-via-malleable-convolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780420.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780420-supp.pdf +tape-task-agnostic-prior-embedding-for-image-restoration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780438.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780438-supp.pdf +uncertainty-inspired-underwater-image-enhancement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780456.pdf, +hourglass-attention-network-for-image-inpainting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780474.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780474-supp.pdf +unfolded-deep-kernel-estimation-for-blind-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780493.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780493-supp.pdf +event-guided-deblurring-of-unknown-exposure-time-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780510.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780510-supp.zip +reconet-recurrent-correction-network-for-fast-and-efficient-multi-modality-image-fusion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780528.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780528-supp.pdf +content-adaptive-latents-and-decoder-for-neural-image-compression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780545.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780545-supp.pdf +efficient-and-degradation-adaptive-network-for-real-world-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780563.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780563-supp.pdf +unidirectional-video-denoising-by-mimicking-backward-recurrent-modules-with-look-ahead-forward-ones,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780581.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780581-supp.pdf +self-supervised-learning-for-real-world-super-resolution-from-dual-zoomed-observations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780599.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780599-supp.pdf +secrets-of-event-based-optical-flow,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780616.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780616-supp.pdf +towards-efficient-and-scale-robust-ultra-high-definition-image-demoireing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780634.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780634-supp.pdf +erdn-equivalent-receptive-field-deformable-network-for-video-deblurring,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780651.pdf, +rethinking-generic-camera-models-for-deep-single-image-camera-calibration-to-recover-rotation-and-fisheye-distortion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780668.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780668-supp.zip +art-ss-an-adaptive-rejection-technique-for-semi-supervised-restoration-for-adverse-weather-affected-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780688.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780688-supp.zip +fusion-from-decomposition-a-self-supervised-decomposition-approach-for-image-fusion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780706.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780706-supp.pdf +learning-degradation-representations-for-image-deblurring,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780724.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136780724-supp.pdf +learning-mutual-modulation-for-self-supervised-cross-modal-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790001-supp.pdf +spectrum-aware-and-transferable-architecture-search-for-hyperspectral-image-restoration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790019-supp.pdf +neural-color-operators-for-sequential-image-retouching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790037.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790037-supp.pdf +optimizing-image-compression-via-joint-learning-with-denoising,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790054.pdf, +restore-globally-refine-locally-a-mask-guided-scheme-to-accelerate-super-resolution-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790072-supp.zip +compiler-aware-neural-architecture-search-for-on-mobile-real-time-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790089.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790089-supp.pdf +modeling-mask-uncertainty-in-hyperspectral-image-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790109.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790109-supp.pdf +perceiving-and-modeling-density-for-image-dehazing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790126.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790126-supp.pdf +stripformer-strip-transformer-for-fast-image-deblurring,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790142.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790142-supp.pdf +deep-fourier-based-exposure-correction-network-with-spatial-frequency-interaction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790159.pdf, +frequency-and-spatial-dual-guidance-for-image-dehazing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790177.pdf, +towards-real-world-hdrtv-reconstruction-a-data-synthesis-based-approach,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790195.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790195-supp.pdf +learning-discriminative-shrinkage-deep-networks-for-image-deconvolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790212.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790212-supp.pdf +kxnet-a-model-driven-deep-neural-network-for-blind-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790230.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790230-supp.pdf +arm-any-time-super-resolution-method,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790248.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790248-supp.pdf +attention-aware-learning-for-hyperparameter-prediction-in-image-processing-pipelines,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790265.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790265-supp.pdf +realflow-em-based-realistic-optical-flow-dataset-generation-from-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790282.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790282-supp.pdf +memory-augmented-model-driven-network-for-pansharpening,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790299.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790299-supp.pdf +all-you-need-is-raw-defending-against-adversarial-attacks-with-camera-image-pipelines,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790316.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790316-supp.pdf +ghost-free-high-dynamic-range-imaging-with-context-aware-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790336.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790336-supp.pdf +style-guided-shadow-removal,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790353.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790353-supp.pdf +d2c-sr-a-divergence-to-convergence-approach-for-real-world-image-super-resolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790370.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790370-supp.pdf +grit-vlp-grouped-mini-batch-sampling-for-efficient-vision-and-language-pre-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790386.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790386-supp.pdf +efficient-video-deblurring-guided-by-motion-magnitude,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790403.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790403-supp.zip +single-frame-atmospheric-turbulence-mitigation-a-benchmark-study-and-a-new-physics-inspired-transformer-model,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790419.pdf, +contextformer-a-transformer-with-spatio-channel-attention-for-context-modeling-in-learned-image-compression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790436.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790436-supp.pdf +image-super-resolution-with-deep-dictionary,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790454.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790454-supp.pdf +tempformer-temporally-consistent-transformer-for-video-denoising,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790471.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790471-supp.zip +rawtobit-a-fully-end-to-end-camera-isp-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790487.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790487-supp.pdf +drcnet-dynamic-image-restoration-contrastive-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790504.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790504-supp.pdf +zero-shot-learning-for-reflection-removal-of-single-360-degree-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790523.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790523-supp.pdf +transformer-with-implicit-edges-for-particle-based-physics-simulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790539.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790539-supp.pdf +rethinking-video-rain-streak-removal-a-new-synthesis-model-and-a-deraining-network-with-video-rain-prior,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790556.pdf, +super-resolution-by-predicting-offsets-an-ultra-efficient-super-resolution-network-for-rasterized-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790572.pdf, +animation-from-blur-multi-modal-blur-decomposition-with-motion-guidance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790588.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790588-supp.zip +alphavc-high-performance-and-efficient-learned-video-compression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790605.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790605-supp.pdf +content-oriented-learned-image-compression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790621.pdf, +rrsr-reciprocal-reference-based-image-super-resolution-with-progressive-feature-alignment-and-selection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790637.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790637-supp.pdf +contrastive-prototypical-network-with-wasserstein-confidence-penalty,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790654.pdf, +learn-to-decompose-cascaded-decomposition-network-for-cross-domain-few-shot-facial-expression-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790672.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790672-supp.pdf +self-support-few-shot-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790689.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790689-supp.pdf +few-shot-object-detection-with-model-calibration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790707.pdf, +self-supervision-can-be-a-good-few-shot-learner,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790726.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136790726-supp.pdf +tsf-transformer-based-semantic-filter-for-few-shot-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800001.pdf, +adversarial-feature-augmentation-for-cross-domain-few-shot-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800019-supp.pdf +constructing-balance-from-imbalance-for-long-tailed-image-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800036-supp.pdf +on-multi-domain-long-tailed-recognition-imbalanced-domain-generalization-and-beyond,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800054.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800054-supp.pdf +few-shot-video-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800071.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800071-supp.pdf +worst-case-matters-for-few-shot-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800092.pdf, +exploring-hierarchical-graph-representation-for-large-scale-zero-shot-image-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800108.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800108-supp.zip +doubly-deformable-aggregation-of-covariance-matrices-for-few-shot-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800125.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800125-supp.pdf +dense-cross-query-and-support-attention-weighted-mask-aggregation-for-few-shot-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800142.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800142-supp.pdf +rethinking-clustering-based-pseudo-labeling-for-unsupervised-meta-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800160.pdf, +claster-clustering-with-reinforcement-learning-for-zero-shot-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800177.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800177-supp.pdf +few-shot-class-incremental-learning-for-3d-point-cloud-objects,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800194.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800194-supp.pdf +meta-learning-with-less-forgetting-on-large-scale-non-stationary-task-distributions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800211.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800211-supp.pdf +dna-improving-few-shot-transfer-learning-with-low-rank-decomposition-and-alignment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800229.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800229-supp.pdf +learning-instance-and-task-aware-dynamic-kernels-for-few-shot-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800247.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800247-supp.pdf +open-world-semantic-segmentation-via-contrasting-and-clustering-vision-language-embedding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800265.pdf, +few-shot-classification-with-contrastive-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800283.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800283-supp.pdf +time-reversed-diffusion-tensor-transformer-a-new-tenet-of-few-shot-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800300.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800300-supp.pdf +self-promoted-supervision-for-few-shot-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800318.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800318-supp.pdf +few-shot-object-counting-and-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800336.pdf, +rethinking-few-shot-object-detection-on-a-multi-domain-benchmark,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800354.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800354-supp.pdf +cross-domain-cross-set-few-shot-learning-via-learning-compact-and-aligned-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800371.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800371-supp.pdf +mutually-reinforcing-structure-with-proposal-contrastive-consistency-for-few-shot-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800388.pdf, +dual-contrastive-learning-with-anatomical-auxiliary-supervision-for-few-shot-medical-image-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800406.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800406-supp.pdf +improving-few-shot-learning-through-multi-task-representation-learning-theory,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800423.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800423-supp.pdf +tree-structure-aware-few-shot-image-classification-via-hierarchical-aggregation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800440.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800440-supp.pdf +inductive-and-transductive-few-shot-video-classification-via-appearance-and-temporal-alignments,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800457.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800457-supp.pdf +temporal-and-cross-modal-attention-for-audio-visual-zero-shot-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800474.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800474-supp.pdf +hm-hybrid-masking-for-few-shot-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800492.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800492-supp.pdf +transvlad-focusing-on-locally-aggregated-descriptors-for-few-shot-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800509.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800509-supp.pdf +kernel-relative-prototype-spectral-filtering-for-few-shot-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800527.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800527-supp.pdf +this-is-my-unicorn-fluffy-personalizing-frozen-vision-language-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800544.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800544-supp.pdf +close-curriculum-learning-on-the-sharing-extent-towards-better-one-shot-nas,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800563.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800563-supp.pdf +streamable-neural-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800580.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800580-supp.zip +gradient-based-uncertainty-for-monocular-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800598.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800598-supp.pdf +online-continual-learning-with-contrastive-vision-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800614.pdf, +cprune-compiler-informed-model-pruning-for-efficient-target-aware-dnn-execution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800634.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800634-supp.pdf +eautodet-efficient-architecture-search-for-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800652.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800652-supp.pdf +a-max-flow-based-approach-for-neural-architecture-search,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800668.pdf, +occamnets-mitigating-dataset-bias-by-favoring-simpler-hypotheses,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800685.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800685-supp.zip +era-enhanced-rational-activations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800705.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800705-supp.pdf +convolutional-embedding-makes-hierarchical-vision-transformer-stronger,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800722.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136800722-supp.pdf +active-label-correction-using-robust-parameter-update-and-entropy-propagation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810001-supp.pdf +unpaired-image-translation-via-vector-symbolic-architectures,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810017.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810017-supp.pdf +uninet-unified-architecture-search-with-convolution-transformer-and-mlp,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810034.pdf, +amixer-adaptive-weight-mixing-for-self-attention-free-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810051.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810051-supp.pdf +tinyvit-fast-pretraining-distillation-for-small-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810068.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810068-supp.pdf +equivariant-hypergraph-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810086.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810086-supp.pdf +scalenet-searching-for-the-model-to-scale,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810103.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810103-supp.pdf +complementing-brightness-constancy-with-deep-networks-for-optical-flow-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810120.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810120-supp.pdf +vitas-vision-transformer-architecture-search,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810138.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810138-supp.pdf +lidarnas-unifying-and-searching-neural-architectures-for-3d-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810156.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810156-supp.pdf +uncertainty-dtw-for-time-series-and-sequences,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810174.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810174-supp.pdf +black-box-few-shot-knowledge-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810191.pdf, +revisiting-batch-norm-initialization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810207.pdf, +ssbnet-improving-visual-recognition-efficiency-by-adaptive-sampling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810224.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810224-supp.pdf +filter-pruning-via-feature-discrimination-in-deep-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810241.pdf, +la3-efficient-label-aware-autoaugment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810258.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810258-supp.pdf +interpretations-steered-network-pruning-via-amortized-inferred-saliency-maps,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810274.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810274-supp.pdf +ba-net-bridge-attention-for-deep-convolutional-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810293.pdf, +sau-smooth-activation-function-using-convolution-with-approximate-identities,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810309.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810309-supp.zip +multi-exit-semantic-segmentation-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810326.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810326-supp.pdf +almost-orthogonal-layers-for-efficient-general-purpose-lipschitz-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810345.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810345-supp.pdf +pointscatter-point-set-representation-for-tubular-structure-extraction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810361.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810361-supp.pdf +check-and-link-pairwise-lesion-correspondence-guides-mammogram-mass-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810379.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810379-supp.pdf +graph-constrained-contrastive-regularization-for-semi-weakly-volumetric-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810396.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810396-supp.pdf +generalizable-medical-image-segmentation-via-random-amplitude-mixup-and-domain-specific-image-restoration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810415.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810415-supp.zip +auto-fedrl-federated-hyperparameter-optimization-for-multi-institutional-medical-image-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810431.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810431-supp.pdf +personalizing-federated-medical-image-segmentation-via-local-calibration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810449.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810449-supp.pdf +one-shot-medical-landmark-localization-by-edge-guided-transform-and-noisy-landmark-refinement,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810466.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810466-supp.pdf +ultra-high-resolution-unpaired-stain-transformation-via-kernelized-instance-normalization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810483.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810483-supp.pdf +med-danet-dynamic-architecture-network-for-efficient-medical-volumetric-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810499.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810499-supp.pdf +concl-concept-contrastive-learning-for-dense-prediction-pre-training-in-pathology-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810516.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810516-supp.pdf +cryoai-amortized-inference-of-poses-for-ab-initio-reconstruction-of-3d-molecular-volumes-from-real-cryo-em-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810533.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810533-supp.pdf +unimiss-universal-medical-self-supervised-learning-via-breaking-dimensionality-barrier,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810551.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810551-supp.pdf +dlme-deep-local-flatness-manifold-embedding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810569.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810569-supp.pdf +semi-supervised-keypoint-detector-and-descriptor-for-retinal-image-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810586.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810586-supp.pdf +graph-neural-network-for-cell-tracking-in-microscopy-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810602.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810602-supp.zip +cxr-segmentation-by-adain-based-domain-adaptation-and-knowledge-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810619.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810619-supp.pdf +accurate-detection-of-proteins-in-cryo-electron-tomograms-from-sparse-labels,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810636.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810636-supp.pdf +k-salsa-k-anonymous-synthetic-averaging-of-retinal-images-via-local-style-alignment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810652.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810652-supp.pdf +radiotransformer-a-cascaded-global-focal-transformer-for-visual-attention-guided-disease-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810669.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810669-supp.pdf +differentiable-zooming-for-multiple-instance-learning-on-whole-slide-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810689.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810689-supp.pdf +learning-uncoupled-modulation-cvae-for-3d-action-conditioned-human-motion-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810707.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810707-supp.zip +towards-grand-unification-of-object-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810724.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136810724-supp.pdf +bytetrack-multi-object-tracking-by-associating-every-detection-box,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820001-supp.pdf +robust-multi-object-tracking-by-marginal-inference,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820020.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820020-supp.pdf +polarmot-how-far-can-geometric-relations-take-us-in-3d-multi-object-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820038.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820038-supp.pdf +particle-video-revisited-tracking-through-occlusions-using-point-trajectories,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820055.pdf, +tracking-objects-as-pixel-wise-distributions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820072-supp.pdf +cmt-context-matching-guided-transformer-for-3d-tracking-in-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820091.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820091-supp.pdf +towards-generic-3d-tracking-in-rgbd-videos-benchmark-and-baseline,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820108.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820108-supp.pdf +hierarchical-latent-structure-for-multi-modal-vehicle-trajectory-forecasting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820125.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820125-supp.pdf +aiatrack-attention-in-attention-for-transformer-visual-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820141.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820141-supp.pdf +disentangling-architecture-and-training-for-optical-flow,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820159.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820159-supp.pdf +a-perturbation-constrained-adversarial-attack-for-evaluating-the-robustness-of-optical-flow,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820177.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820177-supp.pdf +robust-landmark-based-stent-tracking-in-x-ray-fluoroscopy,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820195.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820195-supp.pdf +social-ode-multi-agent-trajectory-forecasting-with-neural-ordinary-differential-equations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820211.pdf, +social-ssl-self-supervised-cross-sequence-representation-learning-based-on-transformers-for-multi-agent-trajectory-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820227.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820227-supp.pdf +diverse-human-motion-prediction-guided-by-multi-level-spatial-temporal-anchors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820244.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820244-supp.pdf +learning-pedestrian-group-representations-for-multi-modal-trajectory-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820263.pdf, +sequential-multi-view-fusion-network-for-fast-lidar-point-motion-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820282.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820282-supp.pdf +e-graph-minimal-solution-for-rigid-rotation-with-extensibility-graphs,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820298.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820298-supp.zip +point-cloud-compression-with-range-image-based-entropy-model-for-autonomous-driving,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820315.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820315-supp.pdf +joint-feature-learning-and-relation-modeling-for-tracking-a-one-stream-framework,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820332.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820332-supp.pdf +motionclip-exposing-human-motion-generation-to-clip-space,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820349.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820349-supp.pdf +backbone-is-all-your-need-a-simplified-architecture-for-visual-object-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820366.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820366-supp.pdf +aware-of-the-history-trajectory-forecasting-with-the-local-behavior-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820383.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820383-supp.pdf +optical-flow-training-under-limited-label-budget-via-active-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820400.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820400-supp.pdf +hierarchical-feature-embedding-for-visual-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820418.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820418-supp.zip +tackling-background-distraction-in-video-object-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820434.pdf, +social-implicit-rethinking-trajectory-prediction-evaluation-and-the-effectiveness-of-implicit-maximum-likelihood-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820451.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820451-supp.pdf +temos-generating-diverse-human-motions-from-textual-descriptions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820468.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820468-supp.pdf +tracking-every-thing-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820486.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820486-supp.pdf +hulc-3d-human-motion-capture-with-pose-manifold-sampling-and-dense-contact-guidance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820503.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820503-supp.zip +towards-sequence-level-training-for-visual-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820521.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820521-supp.pdf +learned-monocular-depth-priors-in-visual-inertial-initialization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820537.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820537-supp.pdf +robust-visual-tracking-by-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820555.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820555-supp.zip +meshloc-mesh-based-visual-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820573.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820573-supp.pdf +s2f2-single-stage-flow-forecasting-for-future-multiple-trajectories-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820593.pdf, +large-displacement-3d-object-tracking-with-hybrid-non-local-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820609.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820609-supp.pdf +fear-fast-efficient-accurate-and-robust-visual-tracker,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820625.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820625-supp.pdf +pref-predictability-regularized-neural-motion-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820643.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820643-supp.zip +view-vertically-a-hierarchical-network-for-trajectory-prediction-via-fourier-spectrums,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820661.pdf, +hvc-net-unifying-homography-visibility-and-confidence-learning-for-planar-object-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820679.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820679-supp.zip +ramgan-region-attentive-morphing-gan-for-region-level-makeup-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820696.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820696-supp.pdf +sinnerf-training-neural-radiance-fields-on-complex-scenes-from-a-single-image,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820712.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820712-supp.pdf +entropy-driven-sampling-and-training-scheme-for-conditional-diffusion-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820730.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136820730-supp.pdf +accelerating-score-based-generative-models-with-preconditioned-diffusion-sampling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830001-supp.pdf +learning-to-generate-realistic-lidar-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830017.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830017-supp.zip +rfnet-4d-joint-object-reconstruction-and-flow-estimation-from-4d-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830036.pdf, +diverse-image-inpainting-with-normalizing-flow,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830053.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830053-supp.pdf +improved-masked-image-generation-with-token-critic,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830070-supp.pdf +trend-truncated-generalized-normal-density-estimation-of-inception-embeddings-for-gan-evaluation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830087.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830087-supp.pdf +exploring-gradient-based-multi-directional-controls-in-gans,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830103.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830103-supp.pdf +spatially-invariant-unsupervised-3d-object-centric-learning-and-scene-decomposition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830120.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830120-supp.pdf +neural-scene-decoration-from-a-single-photograph,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830137.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830137-supp.pdf +outpainting-by-queries,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830154.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830154-supp.pdf +unleashing-transformers-parallel-token-prediction-with-discrete-absorbing-diffusion-for-fast-high-resolution-image-generation-from-vector-quantized-codes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830171.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830171-supp.zip +chunkygan-real-image-inversion-via-segments,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830191.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830191-supp.zip +gan-cocktail-mixing-gans-without-dataset-access,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830207.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830207-supp.pdf +geometry-guided-progressive-nerf-for-generalizable-and-efficient-neural-human-rendering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830224.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830224-supp.zip +controllable-shadow-generation-using-pixel-height-maps,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830240.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830240-supp.pdf +learning-where-to-look-generative-nas-is-surprisingly-efficient,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830257.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830257-supp.pdf +subspace-diffusion-generative-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830274.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830274-supp.pdf +duelgan-a-duel-between-two-discriminators-stabilizes-the-gan-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830290.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830290-supp.zip +miner-multiscale-implicit-neural-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830308.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830308-supp.pdf +an-embedded-feature-whitening-approach-to-deep-neural-network-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830324.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830324-supp.pdf +q-fw-a-hybrid-classical-quantum-frank-wolfe-for-quadratic-binary-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830341.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830341-supp.pdf +self-supervised-learning-of-visual-graph-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830359.pdf, +scalable-learning-to-optimize-a-learned-optimizer-can-train-big-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830376.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830376-supp.pdf +qista-imagenet-a-deep-compressive-image-sensing-framework-solving-lq-norm-optimization-problem,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830394.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830394-supp.pdf +r-dfcil-relation-guided-representation-learning-for-data-free-class-incremental-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830411.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830411-supp.pdf +domain-generalization-by-mutual-information-regularization-with-pre-trained-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830427.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830427-supp.pdf +predicting-is-not-understanding-recognizing-and-addressing-underspecification-in-machine-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830445.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830445-supp.pdf +neural-sim-learning-to-generate-training-data-with-nerf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830463.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830463-supp.pdf +bayesian-optimization-with-clustering-and-rollback-for-cnn-auto-pruning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830480.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830480-supp.pdf +learned-variational-video-color-propagation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830497.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830497-supp.pdf +continual-variational-autoencoder-learning-via-online-cooperative-memorization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830515.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830515-supp.pdf +learning-to-learn-with-smooth-regularization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830533.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830533-supp.pdf +incremental-task-learning-with-incremental-rank-updates,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830549.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830549-supp.pdf +batch-efficient-eigendecomposition-for-small-and-medium-matrices,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830566.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830566-supp.pdf +ensemble-learning-priors-driven-deep-unfolding-for-scalable-video-snapshot-compressive-imaging,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830583.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830583-supp.zip +approximate-discrete-optimal-transport-plan-with-auxiliary-measure-method,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830602.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830602-supp.pdf +a-comparative-study-of-graph-matching-algorithms-in-computer-vision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830618.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830618-supp.pdf +improving-generalization-in-federated-learning-by-seeking-flat-minima,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830636.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830636-supp.pdf +semidefinite-relaxations-of-truncated-least-squares-in-robust-rotation-search-tight-or-not,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830655.pdf, +transfer-without-forgetting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830672.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830672-supp.pdf +adabest-minimizing-client-drift-in-federated-learning-via-adaptive-bias-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830690.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830690-supp.pdf +tackling-long-tailed-category-distribution-under-domain-shifts,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830706.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830706-supp.pdf +doubly-fused-vit-fuse-information-from-vision-transformer-doubly-with-local-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830723.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136830723-supp.pdf +improving-vision-transformers-by-revisiting-high-frequency-components,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840001-supp.pdf +recurrent-bilinear-optimization-for-binary-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840019.pdf, +neural-architecture-search-for-spiking-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840036-supp.pdf +where-to-focus-investigating-hierarchical-attention-relationship-for-fine-grained-visual-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840056.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840056-supp.pdf +davit-dual-attention-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840073.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840073-supp.pdf +optimal-transport-for-label-efficient-visible-infrared-person-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840091.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840091-supp.pdf +locality-guidance-for-improving-vision-transformers-on-tiny-datasets,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840108.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840108-supp.pdf +neighborhood-collective-estimation-for-noisy-label-identification-and-correction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840126.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840126-supp.pdf +few-shot-class-incremental-learning-via-entropy-regularized-data-free-replay,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840144.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840144-supp.pdf +anti-retroactive-interference-for-lifelong-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840160.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840160-supp.pdf +towards-calibrated-hyper-sphere-representation-via-distribution-overlap-coefficient-for-long-tailed-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840176.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840176-supp.pdf +dynamic-metric-learning-with-cross-level-concept-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840194.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840194-supp.pdf +menet-a-memory-based-network-with-dual-branch-for-efficient-event-stream-processing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840211.pdf, +out-of-distribution-detection-with-boundary-aware-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840232.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840232-supp.pdf +learning-hierarchy-aware-features-for-reducing-mistake-severity,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840249.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840249-supp.pdf +learning-to-detect-every-thing-in-an-open-world,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840265.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840265-supp.pdf +kvt-k-nn-attention-for-boosting-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840281.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840281-supp.pdf +registration-based-few-shot-anomaly-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840300.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840300-supp.pdf +improving-robustness-by-enhancing-weak-subnets,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840317.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840317-supp.pdf +learning-invariant-visual-representations-for-compositional-zero-shot-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840335.pdf, +improving-covariance-conditioning-of-the-svd-meta-layer-by-orthogonality,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840352.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840352-supp.pdf +out-of-distribution-detection-with-semantic-mismatch-under-masking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840369.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840369-supp.pdf +data-free-neural-architecture-search-via-recursive-label-calibration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840386.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840386-supp.pdf +learning-from-multiple-annotator-noisy-labels-via-sample-wise-label-fusion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840402.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840402-supp.pdf +acknowledging-the-unknown-for-multi-label-learning-with-single-positive-labels,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840418.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840418-supp.pdf +automix-unveiling-the-power-of-mixup-for-stronger-classifiers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840435.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840435-supp.pdf +maxvit-multi-axis-vision-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840453.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840453-supp.pdf +scalablevit-rethinking-the-context-oriented-generalization-of-vision-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840473.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840473-supp.pdf +three-things-everyone-should-know-about-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840490.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840490-supp.pdf +deit-iii-revenge-of-the-vit,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840509.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840509-supp.pdf +mixskd-self-knowledge-distillation-from-mixup-for-image-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840527.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840527-supp.pdf +self-feature-distillation-with-uncertainty-modeling-for-degraded-image-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840544.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840544-supp.pdf +novel-class-discovery-without-forgetting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840561.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840561-supp.pdf +safa-sample-adaptive-feature-augmentation-for-long-tailed-image-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840578.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840578-supp.pdf +negative-samples-are-at-large-leveraging-hard-distance-elastic-loss-for-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840595.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840595-supp.pdf +discrete-constrained-regression-for-local-counting-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840612.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840612-supp.pdf +breadcrumbs-adversarial-class-balanced-sampling-for-long-tailed-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840628.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840628-supp.pdf +chairs-can-be-stood-on-overcoming-object-bias-in-human-object-interaction-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840645.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840645-supp.pdf +a-fast-knowledge-distillation-framework-for-visual-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840663.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840663-supp.pdf +dice-leveraging-sparsification-for-out-of-distribution-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840680.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840680-supp.pdf +invariant-feature-learning-for-generalized-long-tailed-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840698.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840698-supp.pdf +sliced-recursive-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840716.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136840716-supp.pdf +cross-domain-ensemble-distillation-for-domain-generalization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850001-supp.pdf +centrality-and-consistency-two-stage-clean-samples-identification-for-learning-with-instance-dependent-noisy-labels,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850021.pdf, +hyperspherical-learning-in-multi-label-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850038.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850038-supp.pdf +when-active-learning-meets-implicit-semantic-data-augmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850056.pdf, +vl-ltr-learning-class-wise-visual-linguistic-representation-for-long-tailed-visual-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850072-supp.pdf +class-is-invariant-to-context-and-vice-versa-on-learning-invariance-for-out-of-distribution-generalization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850089.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850089-supp.pdf +hierarchical-semi-supervised-contrastive-learning-for-contamination-resistant-anomaly-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850107.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850107-supp.pdf +tracking-by-associating-clips,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850126.pdf, +realpatch-a-statistical-matching-framework-for-model-patching-with-real-samples,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850144.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850144-supp.pdf +background-insensitive-scene-text-recognition-with-text-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850161.pdf, +semantic-novelty-detection-via-relational-reasoning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850181.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850181-supp.pdf +improving-closed-and-open-vocabulary-attribute-prediction-using-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850199.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850199-supp.pdf +training-vision-transformers-with-only-2040-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850218.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850218-supp.pdf +bridging-images-and-videos-a-simple-learning-framework-for-large-vocabulary-video-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850235.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850235-supp.pdf +tdam-top-down-attention-module-for-contextually-guided-feature-selection-in-cnns,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850255.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850255-supp.pdf +automatic-check-out-via-prototype-based-classifier-learning-from-single-product-exemplars,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850273.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850273-supp.pdf +overcoming-shortcut-learning-in-a-target-domain-by-generalizing-basic-visual-factors-from-a-source-domain,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850290.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850290-supp.pdf +photo-realistic-neural-domain-randomization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850306.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850306-supp.zip +wave-vit-unifying-wavelet-and-transformers-for-visual-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850324.pdf, +tailoring-self-supervision-for-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850342.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850342-supp.pdf +difficulty-aware-simulator-for-open-set-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850360.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850360-supp.pdf +few-shot-class-incremental-learning-from-an-open-set-perspective,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850377.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850377-supp.pdf +foster-feature-boosting-and-compression-for-class-incremental-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850393.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850393-supp.pdf +visual-knowledge-tracing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850410.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850410-supp.pdf +s3c-self-supervised-stochastic-classifiers-for-few-shot-class-incremental-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850427.pdf, +improving-fine-grained-visual-recognition-in-low-data-regimes-via-self-boosting-attention-mechanism,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850444.pdf, +vsa-learning-varied-size-window-attention-in-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850460.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850460-supp.pdf +unbiased-manifold-augmentation-for-coarse-class-subdivision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850478.pdf, +densehybrid-hybrid-anomaly-detection-for-dense-open-set-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850494.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850494-supp.pdf +rethinking-confidence-calibration-for-failure-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850512.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850512-supp.pdf +uncertainty-guided-source-free-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850530.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850530-supp.pdf +should-all-proposals-be-treated-equally-in-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850549.pdf, +vip-unified-certified-detection-and-recovery-for-patch-attack-with-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850566.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850566-supp.pdf +incdfm-incremental-deep-feature-modeling-for-continual-novelty-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850581.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850581-supp.pdf +igformer-interaction-graph-transformer-for-skeleton-based-human-interaction-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850598.pdf, +prime-a-few-primitives-can-boost-robustness-to-common-corruptions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850615.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850615-supp.pdf +rotation-regularization-without-rotation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850632.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850632-supp.pdf +towards-accurate-open-set-recognition-via-background-class-regularization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850648.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850648-supp.pdf +in-defense-of-image-pre-training-for-spatiotemporal-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850665.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850665-supp.pdf +augmenting-deep-classifiers-with-polynomial-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850682.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850682-supp.pdf +learning-with-noisy-labels-by-efficient-transition-matrix-estimation-to-combat-label-miscorrection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850700.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850700-supp.pdf +online-task-free-continual-learning-with-dynamic-sparse-distributed-memory,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136850721.pdf, +contrastive-deep-supervision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860001.pdf, +discriminability-transferability-trade-off-an-information-theoretic-perspective,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860020.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860020-supp.pdf +locvtp-video-text-pre-training-for-temporal-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860037.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860037-supp.pdf +few-shot-end-to-end-object-detection-via-constantly-concentrated-encoding-across-heads,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860056.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860056-supp.pdf +implicit-neural-representations-for-image-compression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860073.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860073-supp.pdf +lip-flow-learning-inference-time-priors-for-codec-avatars-via-normalizing-flows-in-latent-space,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860091.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860091-supp.pdf +learning-to-drive-by-watching-youtube-videos-action-conditioned-contrastive-policy-pretraining,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860109.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860109-supp.pdf +learning-ego-3d-representation-as-ray-tracing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860126.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860126-supp.pdf +static-and-dynamic-concepts-for-self-supervised-video-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860142.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860142-supp.pdf +spherefed-hyperspherical-federated-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860161.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860161-supp.pdf +hierarchically-self-supervised-transformer-for-human-skeleton-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860181.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860181-supp.pdf +posterior-refinement-on-metric-matrix-improves-generalization-bound-in-metric-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860199.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860199-supp.pdf +balancing-stability-and-plasticity-through-advanced-null-space-in-continual-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860215.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860215-supp.pdf +disco-remedying-self-supervised-learning-on-lightweight-models-with-distilled-contrastive-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860233.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860233-supp.pdf +coscl-cooperation-of-small-continual-learners-is-stronger-than-a-big-one,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860249.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860249-supp.pdf +manifold-adversarial-learning-for-cross-domain-3d-shape-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860266.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860266-supp.pdf +fast-moco-boost-momentum-based-contrastive-learning-with-combinatorial-patches,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860283.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860283-supp.pdf +lord-local-4d-implicit-representation-for-high-fidelity-dynamic-human-modeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860299.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860299-supp.pdf +on-the-versatile-uses-of-partial-distance-correlation-in-deep-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860318.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860318-supp.pdf +self-regulated-feature-learning-via-teacher-free-feature-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860337.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860337-supp.pdf +balancing-between-forgetting-and-acquisition-in-incremental-subpopulation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860354.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860354-supp.pdf +counterfactual-intervention-feature-transfer-for-visible-infrared-person-re-identification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860371.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860371-supp.pdf +das-densely-anchored-sampling-for-deep-metric-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860388.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860388-supp.pdf +learn-from-all-erasing-attention-consistency-for-noisy-label-facial-expression-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860406.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860406-supp.pdf +a-non-isotropic-probabilistic-take-on-proxy-based-deep-metric-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860423.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860423-supp.pdf +tokenmix-rethinking-image-mixing-for-data-augmentation-in-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860442.pdf, +ufo-unified-feature-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860459.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860459-supp.pdf +sound-localization-by-self-supervised-time-delay-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860476.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860476-supp.pdf +x-learner-learning-cross-sources-and-tasks-for-universal-visual-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860495.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860495-supp.pdf +slip-self-supervision-meets-language-image-pre-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860514.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860514-supp.pdf +discovering-deformable-keypoint-pyramids,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860531.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860531-supp.pdf +neural-video-compression-using-gans-for-detail-synthesis-and-propagation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860549.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860549-supp.pdf +a-contrastive-objective-for-learning-disentangled-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860566.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860566-supp.pdf +pt4al-using-self-supervised-pretext-tasks-for-active-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860583.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860583-supp.pdf +parc-net-position-aware-circular-convolution-with-merits-from-convnets-and-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860600.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860600-supp.pdf +dualprompt-complementary-prompting-for-rehearsal-free-continual-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860617.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860617-supp.pdf +unifying-visual-contrastive-learning-for-object-recognition-from-a-graph-perspective,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860635.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860635-supp.pdf +decoupled-contrastive-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860653.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860653-supp.pdf +joint-learning-of-localized-representations-from-medical-images-and-reports,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860670.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860670-supp.pdf +the-challenges-of-continuous-self-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860687.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860687-supp.pdf +conditional-stroke-recovery-for-fine-grained-sketch-based-image-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860708.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860708-supp.pdf +identifying-hard-noise-in-long-tailed-sample-distribution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860725.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136860725-supp.pdf +relative-contrastive-loss-for-unsupervised-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870001-supp.pdf +fine-grained-fashion-representation-learning-by-online-deep-clustering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870019-supp.pdf +nashae-disentangling-representations-through-adversarial-covariance-minimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870036-supp.pdf +a-gyrovector-space-approach-for-symmetric-positive-semi-definite-matrix-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870052.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870052-supp.pdf +learning-visual-representation-from-modality-shared-contrastive-language-image-pre-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870069.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870069-supp.pdf +contrasting-quadratic-assignments-for-set-based-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870087.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870087-supp.pdf +class-incremental-learning-with-cross-space-clustering-and-controlled-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870104.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870104-supp.pdf +object-discovery-and-representation-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870121.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870121-supp.pdf +trading-positional-complexity-vs-deepness-in-coordinate-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870142.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870142-supp.pdf +mvdg-a-unified-multi-view-framework-for-domain-generalization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870158.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870158-supp.pdf +panoptic-scene-graph-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870175.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870175-supp.pdf +object-compositional-neural-implicit-surfaces,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870194.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870194-supp.pdf +rignet-repetitive-image-guided-network-for-depth-completion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870211.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870211-supp.pdf +fade-fusing-the-assets-of-decoder-and-encoder-for-task-agnostic-upsampling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870228.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870228-supp.pdf +lidal-inter-frame-uncertainty-based-active-learning-for-3d-lidar-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870245.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870245-supp.pdf +hierarchical-memory-learning-for-fine-grained-scene-graph-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870263.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870263-supp.pdf +doda-data-oriented-sim-to-real-domain-adaptation-for-3d-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870280.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870280-supp.pdf +mtformer-multi-task-learning-via-transformer-and-cross-task-reasoning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870299.pdf, +monoplflownet-permutohedral-lattice-flownet-for-real-scale-3d-scene-flow-estimation-with-monocular-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870316.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870316-supp.pdf +to-scene-a-large-scale-dataset-for-understanding-3d-tabletop-scenes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870334.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870334-supp.pdf +is-it-necessary-to-transfer-temporal-knowledge-for-domain-adaptive-video-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870351.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870351-supp.zip +meta-spatio-temporal-debiasing-for-video-scene-graph-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870368.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870368-supp.pdf +improving-the-reliability-for-confidence-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870385.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870385-supp.pdf +fine-grained-scene-graph-generation-with-data-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870402.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870402-supp.pdf +pose2room-understanding-3d-scenes-from-human-activities,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870418.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870418-supp.zip +towards-hard-positive-query-mining-for-detr-based-human-object-interaction-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870437.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870437-supp.pdf +discovering-human-object-interaction-concepts-via-self-compositional-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870454.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870454-supp.pdf +primitive-based-shape-abstraction-via-nonparametric-bayesian-inference,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870472.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870472-supp.pdf +stereo-depth-estimation-with-echoes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870489.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870489-supp.pdf +inverted-pyramid-multi-task-transformer-for-dense-scene-understanding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870506.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870506-supp.pdf +petr-position-embedding-transformation-for-multi-view-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870523.pdf, +s2net-stochastic-sequential-pointcloud-forecasting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870541.pdf, +ra-depth-resolution-adaptive-self-supervised-monocular-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870557.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870557-supp.pdf +polyphonicformer-unified-query-learning-for-depth-aware-video-panoptic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870574.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870574-supp.pdf +sqn-weakly-supervised-semantic-segmentation-of-large-scale-3d-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870592.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870592-supp.pdf +pointmixer-mlp-mixer-for-point-cloud-understanding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870611.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870611-supp.pdf +initialization-and-alignment-for-adversarial-texture-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870631.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870631-supp.pdf +motr-end-to-end-multiple-object-tracking-with-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870648.pdf, +gala-toward-geometry-and-lighting-aware-object-search-for-compositing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870665.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870665-supp.pdf +lalaloc-global-floor-plan-comprehension-for-layout-localisation-in-unvisited-environments,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870681.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870681-supp.pdf +3d-pl-domain-adaptive-depth-estimation-with-3d-aware-pseudo-labeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870698.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870698-supp.pdf +panoptic-partformer-learning-a-unified-model-for-panoptic-part-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870716.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136870716-supp.pdf +salient-object-detection-for-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880001.pdf, +learning-semantic-segmentation-from-multiple-datasets-with-label-shifts,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880019-supp.pdf +weakly-supervised-3d-scene-segmentation-with-region-level-boundary-awareness-and-instance-discrimination,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880036-supp.pdf +towards-open-vocabulary-scene-graph-generation-with-prompt-based-finetuning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880055.pdf, +variance-aware-weight-initialization-for-point-convolutional-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880073.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880073-supp.pdf +break-and-make-interactive-structural-understanding-using-lego-bricks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880089.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880089-supp.zip +bi-pointflownet-bidirectional-learning-for-point-cloud-based-scene-flow-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880107.pdf, +3dg-stfm-3d-geometric-guided-student-teacher-feature-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880124.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880124-supp.zip +video-restoration-framework-and-its-meta-adaptations-to-data-poor-conditions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880142.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880142-supp.pdf +monteboxfinder-detecting-and-filtering-primitives-to-fit-a-noisy-point-cloud,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880160.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880160-supp.zip +scene-text-recognition-with-permuted-autoregressive-sequence-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880177.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880177-supp.pdf +when-counting-meets-hmer-counting-aware-network-for-handwritten-mathematical-expression-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880197.pdf, +detecting-tampered-scene-text-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880214.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880214-supp.pdf +optimal-boxes-boosting-end-to-end-scene-text-recognition-by-adjusting-annotated-bounding-boxes-via-reinforcement-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880231.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880231-supp.pdf +glass-global-to-local-attention-for-scene-text-spotting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880248.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880248-supp.pdf +coo-comic-onomatopoeia-dataset-for-recognizing-arbitrary-or-truncated-texts,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880265.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880265-supp.pdf +language-matters-a-weakly-supervised-vision-language-pre-training-approach-for-scene-text-detection-and-spotting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880282.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880282-supp.pdf +toward-understanding-wordart-corner-guided-transformer-for-scene-text-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880301.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880301-supp.pdf +levenshtein-ocr,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880319.pdf, +multi-granularity-prediction-for-scene-text-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880336.pdf, +dynamic-low-resolution-distillation-for-cost-efficient-end-to-end-text-spotting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880353.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880353-supp.pdf +contextual-text-block-detection-towards-scene-text-understanding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880371.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880371-supp.pdf +comer-modeling-coverage-for-transformer-based-handwritten-mathematical-expression-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880389.pdf, +dont-forget-me-accurate-background-recovery-for-text-removal-via-modeling-local-global-context,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880406.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880406-supp.pdf +textadain-paying-attention-to-shortcut-learning-in-text-recognizers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880423.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880423-supp.pdf +multi-modal-text-recognition-networks-interactive-enhancements-between-visual-and-semantic-features,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880442.pdf, +sgbanet-semantic-gan-and-balanced-attention-network-for-arbitrarily-oriented-scene-text-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880459.pdf, +pure-transformer-with-integrated-experts-for-scene-text-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880476.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880476-supp.pdf +ocr-free-document-understanding-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880493.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880493-supp.pdf +car-class-aware-regularizations-for-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880514.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880514-supp.pdf +style-hallucinated-dual-consistency-learning-for-domain-generalized-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880530.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880530-supp.pdf +seqformer-sequential-transformer-for-video-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880547.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880547-supp.pdf +saliency-hierarchy-modeling-via-generative-kernels-for-salient-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880564.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880564-supp.pdf +in-defense-of-online-models-for-video-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880582.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880582-supp.pdf +active-pointly-supervised-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880599.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880599-supp.pdf +a-transformer-based-decoder-for-semantic-segmentation-with-multi-level-context-mining,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880617.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880617-supp.pdf +xmem-long-term-video-object-segmentation-with-an-atkinson-shiffrin-memory-model,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880633.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880633-supp.pdf +self-distillation-for-robust-lidar-semantic-segmentation-in-autonomous-driving,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880650.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880650-supp.pdf +2dpass-2d-priors-assisted-semantic-segmentation-on-lidar-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880668.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880668-supp.pdf +extract-free-dense-labels-from-clip,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880687.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880687-supp.pdf +3d-compositional-zero-shot-learning-with-decompositional-consensus,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880704.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880704-supp.pdf +video-mask-transfiner-for-high-quality-video-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880721.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880721-supp.pdf +box-supervised-instance-segmentation-with-level-set-evolution,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890001.pdf, +point-primitive-transformer-for-long-term-4d-point-cloud-video-understanding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890018.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890018-supp.pdf +adaptive-agent-transformer-for-few-shot-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890035.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890035-supp.zip +waymo-open-dataset-panoramic-video-panoptic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890052.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890052-supp.zip +transfgu-a-top-down-approach-to-fine-grained-unsupervised-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890072-supp.pdf +adaafford-learning-to-adapt-manipulation-affordance-for-3d-articulated-objects-via-few-shot-interactions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890089.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890089-supp.zip +cost-aggregation-with-4d-convolutional-swin-transformer-for-few-shot-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890106.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890106-supp.pdf +fine-grained-egocentric-hand-object-segmentation-dataset-model-and-applications,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890125.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890125-supp.zip +perceptual-artifacts-localization-for-inpainting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890145.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890145-supp.pdf +2d-amodal-instance-segmentation-guided-by-3d-shape-prior,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890164.pdf, +data-efficient-3d-learner-via-knowledge-transferred-from-2d-model,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890181.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890181-supp.pdf +adaptive-spatial-bce-loss-for-weakly-supervised-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890198.pdf, +dense-gaussian-processes-for-few-shot-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890215.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890215-supp.pdf +3d-instances-as-1d-kernels,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890233.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890233-supp.pdf +transmatting-enhancing-transparent-objects-matting-with-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890250.pdf, +mvsalnet-multi-view-augmentation-for-rgb-d-salient-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890268.pdf, +k-means-mask-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890286.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890286-supp.pdf +segpgd-an-effective-and-efficient-adversarial-attack-for-evaluating-and-boosting-segmentation-robustness,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890306.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890306-supp.pdf +adversarial-erasing-framework-via-triplet-with-gated-pyramid-pooling-layer-for-weakly-supervised-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890323.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890323-supp.pdf +continual-semantic-segmentation-via-structure-preserving-and-projected-feature-alignment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890341.pdf, +interclass-prototype-relation-for-few-shot-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890358.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890358-supp.pdf +slim-scissors-segmenting-thin-object-from-synthetic-background,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890375.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890375-supp.pdf +abstracting-sketches-through-simple-primitives,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890392.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890392-supp.pdf +multi-scale-and-cross-scale-contrastive-learning-for-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890408.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890408-supp.pdf +one-trimap-video-matting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890426.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890426-supp.pdf +d2ada-dynamic-density-aware-active-domain-adaptation-for-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890443.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890443-supp.pdf +learning-quality-aware-dynamic-memory-for-video-object-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890462.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890462-supp.pdf +learning-implicit-feature-alignment-function-for-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890479.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890479-supp.pdf +quantum-motion-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890497.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890497-supp.pdf +instance-as-identity-a-generic-online-paradigm-for-video-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890515.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890515-supp.zip +laplacian-mesh-transformer-dual-attention-and-topology-aware-network-for-3d-mesh-classification-and-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890532.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890532-supp.pdf +geodesic-former-a-geodesic-guided-few-shot-3d-point-cloud-instance-segmenter,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890552.pdf, +union-set-multi-source-model-adaptation-for-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890570.pdf, +point-mixswap-attentional-point-cloud-mixing-via-swapping-matched-structural-divisions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890587.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890587-supp.zip +batman-bilateral-attention-transformer-in-motion-appearance-neighboring-space-for-video-object-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890603.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890603-supp.pdf +spsn-superpixel-prototype-sampling-network-for-rgb-d-salient-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890621.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890621-supp.pdf +global-spectral-filter-memory-network-for-video-object-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890639.pdf, +video-instance-segmentation-via-multi-scale-spatio-temporal-split-attention-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890657.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890657-supp.pdf +rankseg-adaptive-pixel-classification-with-image-category-ranking-for-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890673.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890673-supp.pdf +learning-topological-interactions-for-multi-class-medical-image-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890691.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890691-supp.pdf +unsupervised-segmentation-in-real-world-images-via-spelke-object-inference,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890708.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890708-supp.pdf +a-simple-baseline-for-open-vocabulary-semantic-segmentation-with-pre-trained-vision-language-model,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890725.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136890725-supp.pdf +fast-two-view-motion-segmentation-using-christoffel-polynomials,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900001-supp.pdf +uctnet-uncertainty-aware-cross-modal-transformer-network-for-indoor-rgb-d-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900020.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900020-supp.pdf +bi-directional-contrastive-learning-for-domain-adaptive-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900038.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900038-supp.pdf +learning-regional-purity-for-instance-segmentation-on-3d-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900055.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900055-supp.pdf +cross-domain-few-shot-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900072-supp.pdf +generative-subgraph-contrast-for-self-supervised-graph-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900090.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900090-supp.pdf +sdae-self-distillated-masked-autoencoder,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900107.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900107-supp.pdf +demystifying-unsupervised-semantic-correspondence-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900124.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900124-supp.pdf +open-set-semi-supervised-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900142.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900142-supp.pdf +vibration-based-uncertainty-estimation-for-learning-from-limited-supervision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900160.pdf, +concurrent-subsidiary-supervision-for-unsupervised-source-free-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900177.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900177-supp.pdf +weakly-supervised-object-localization-through-inter-class-feature-similarity-and-intra-class-appearance-consistency,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900194.pdf, +active-learning-strategies-for-weakly-supervised-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900210.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900210-supp.pdf +mc-beit-multi-choice-discretization-for-image-bert-pre-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900229.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900229-supp.pdf +bootstrapped-masked-autoencoders-for-vision-bert-pretraining,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900246.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900246-supp.pdf +unsupervised-visual-representation-learning-by-synchronous-momentum-grouping,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900264.pdf, +improving-few-shot-part-segmentation-using-coarse-supervision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900282.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900282-supp.pdf +what-to-hide-from-your-students-attention-guided-masked-image-modeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900299.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900299-supp.pdf +pointly-supervised-panoptic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900318.pdf, +mvp-multimodality-guided-visual-pre-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900336.pdf, +locally-varying-distance-transform-for-unsupervised-visual-anomaly-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900353.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900353-supp.pdf +hrda-context-aware-high-resolution-domain-adaptive-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900370.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900370-supp.pdf +spot-the-difference-self-supervised-pre-training-for-anomaly-detection-and-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900389.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900389-supp.pdf +dual-domain-self-supervised-learning-and-model-adaption-for-deep-compressive-imaging,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900406.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900406-supp.pdf +unsupervised-selective-labeling-for-more-effective-semi-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900423.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900423-supp.pdf +max-pooling-with-vision-transformers-reconciles-class-and-shape-in-weakly-supervised-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900442.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900442-supp.pdf +dense-siamese-network-for-dense-unsupervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900460.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900460-supp.pdf +multi-granularity-distillation-scheme-towards-lightweight-semi-supervised-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900477.pdf, +cp2-copy-paste-contrastive-pretraining-for-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900494.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900494-supp.pdf +self-filtering-a-noise-aware-sample-selection-for-label-noise-with-confidence-penalization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900511.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900511-supp.pdf +rda-reciprocal-distribution-alignment-for-robust-semi-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900527.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900527-supp.pdf +memsac-memory-augmented-sample-consistency-for-large-scale-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900543.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900543-supp.pdf +united-defocus-blur-detection-and-deblurring-via-adversarial-promoting-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900562.pdf, +synergistic-self-supervised-and-quantization-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900579.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900579-supp.pdf +semi-supervised-vision-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900596.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900596-supp.pdf +domain-adaptive-video-segmentation-via-temporal-pseudo-supervision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900612.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900612-supp.pdf +diverse-learner-exploring-diverse-supervision-for-semi-supervised-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900631.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900631-supp.pdf +a-closer-look-at-invariances-in-self-supervised-pre-training-for-3d-vision,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900647.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900647-supp.pdf +conmatch-semi-supervised-learning-with-confidence-guided-consistency-regularization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900665.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900665-supp.pdf +fedx-unsupervised-federated-learning-with-cross-knowledge-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900682.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900682-supp.pdf +w2n-switching-from-weak-supervision-to-noisy-supervision-for-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900699.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900699-supp.pdf +decoupled-adversarial-contrastive-learning-for-self-supervised-adversarial-robustness,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900716.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136900716-supp.pdf +goca-guided-online-cluster-assignment-for-self-supervised-video-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910001.pdf, +constrained-mean-shift-using-distant-yet-related-neighbors-for-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910021.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910021-supp.pdf +revisiting-the-critical-factors-of-augmentation-invariant-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910040.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910040-supp.pdf +ca-ssl-class-agnostic-semi-supervised-learning-for-detection-and-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910057.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910057-supp.pdf +dual-adaptive-transformations-for-weakly-supervised-point-cloud-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910075.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910075-supp.pdf +semantic-aware-fine-grained-correspondence,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910093.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910093-supp.zip +self-supervised-classification-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910112.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910112-supp.pdf +data-invariants-to-understand-unsupervised-out-of-distribution-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910129.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910129-supp.pdf +domain-invariant-masked-autoencoders-for-self-supervised-learning-from-multi-domains,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910147.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910147-supp.pdf +semi-supervised-object-detection-via-virtual-category-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910164.pdf, +completely-self-supervised-crowd-counting-via-distribution-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910180.pdf, +coarse-to-fine-incremental-few-shot-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910199.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910199-supp.pdf +learning-unbiased-transferability-for-domain-adaptation-by-uncertainty-modeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910216.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910216-supp.pdf +learn2augment-learning-to-composite-videos-for-data-augmentation-in-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910234.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910234-supp.pdf +cyborgs-contrastively-bootstrapping-object-representations-by-grounding-in-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910251.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910251-supp.pdf +pss-progressive-sample-selection-for-open-world-visual-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910269.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910269-supp.pdf +improving-self-supervised-lightweight-model-learning-via-hard-aware-metric-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910286.pdf, +object-discovery-via-contrastive-learning-for-weakly-supervised-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910302.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910302-supp.pdf +stochastic-consensus-enhancing-semi-supervised-learning-with-consistency-of-stochastic-classifiers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910319.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910319-supp.pdf +diffusemorph-unsupervised-deformable-image-registration-using-diffusion-model,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910336.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910336-supp.pdf +semi-leak-membership-inference-attacks-against-semi-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910353.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910353-supp.pdf +openldn-learning-to-discover-novel-classes-for-open-world-semi-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910370.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910370-supp.pdf +embedding-contrastive-unsupervised-features-to-cluster-in-and-out-of-distribution-noise-in-corrupted-image-datasets,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910389.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910389-supp.pdf +unsupervised-few-shot-image-classification-by-learning-features-into-clustering-space,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910406.pdf, +towards-realistic-semi-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910423.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910423-supp.pdf +masked-siamese-networks-for-label-efficient-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910442.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910442-supp.pdf +natural-synthetic-anomalies-for-self-supervised-anomaly-detection-and-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910459.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910459-supp.pdf +understanding-collapse-in-non-contrastive-siamese-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910476.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910476-supp.pdf +federated-self-supervised-learning-for-video-understanding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910492.pdf, +towards-efficient-and-effective-self-supervised-learning-of-visual-representations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910509.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910509-supp.pdf +dsr-a-dual-subspace-re-projection-network-for-surface-anomaly-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910526.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910526-supp.pdf +pseudoaugment-learning-to-use-unlabeled-data-for-data-augmentation-in-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910542.pdf, +mvster-epipolar-transformer-for-efficient-multi-view-stereo,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910561.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910561-supp.pdf +relpose-predicting-probabilistic-relative-rotation-for-single-objects-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910580.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910580-supp.pdf +r2l-distilling-neural-radiance-field-to-neural-light-field-for-efficient-novel-view-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910598.pdf, +kd-mvs-knowledge-distillation-based-self-supervised-learning-for-multi-view-stereo,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910615.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910615-supp.pdf +salve-semantic-alignment-verification-for-floorplan-reconstruction-from-sparse-panoramas,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910632.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910632-supp.pdf +rc-mvsnet-unsupervised-multi-view-stereo-with-neural-rendering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910649.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910649-supp.zip +box2mask-weakly-supervised-3d-semantic-instance-segmentation-using-bounding-boxes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910666.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910666-supp.pdf +neilf-neural-incident-light-field-for-physically-based-material-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910684.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910684-supp.zip +arf-artistic-radiance-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910701.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910701-supp.pdf +multiview-stereo-with-cascaded-epipolar-raft,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910718.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136910718-supp.pdf +arah-animatable-volume-rendering-of-articulated-human-sdfs,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920001-supp.pdf +aspanformer-detector-free-image-matching-with-adaptive-span-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920020.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920020-supp.pdf +ndf-neural-deformable-fields-for-dynamic-human-modelling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920037.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920037-supp.pdf +neural-density-distance-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920053.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920053-supp.zip +next-towards-high-quality-neural-radiance-fields-via-multi-skip-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920069.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920069-supp.pdf +learning-online-multi-sensor-depth-fusion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920088.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920088-supp.pdf +bungeenerf-progressive-neural-radiance-field-for-extreme-multi-scale-scene-rendering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920106.pdf, +decomposing-the-tangent-of-occluding-boundaries-according-to-curvatures-and-torsions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920123.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920123-supp.pdf +neuris-neural-reconstruction-of-indoor-scenes-using-normal-priors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920139.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920139-supp.pdf +generalizable-patch-based-neural-rendering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920156.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920156-supp.pdf +improving-rgb-d-point-cloud-registration-by-learning-multi-scale-local-linear-transformation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920175.pdf, +real-time-neural-character-rendering-with-pose-guided-multiplane-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920192.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920192-supp.pdf +sparseneus-fast-generalizable-neural-surface-reconstruction-from-sparse-views,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920210.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920210-supp.pdf +disentangling-object-motion-and-occlusion-for-unsupervised-multi-frame-monocular-depth,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920228.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920228-supp.pdf +depth-field-networks-for-generalizable-multi-view-scene-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920245.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920245-supp.zip +context-enhanced-stereo-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920263.pdf, +pcw-net-pyramid-combination-and-warping-cost-volume-for-stereo-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920280.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920280-supp.pdf +gen6d-generalizable-model-free-6-dof-object-pose-estimation-from-rgb-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920297.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920297-supp.pdf +latency-aware-collaborative-perception,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920315.pdf, +tensorf-tensorial-radiance-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920332.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920332-supp.pdf +nefsac-neurally-filtered-minimal-samples,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920350.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920350-supp.pdf +snes-learning-probably-symmetric-neural-surfaces-from-incomplete-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920366.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920366-supp.zip +hdr-plenoxels-self-calibrating-high-dynamic-range-radiance-fields,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920383.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920383-supp.pdf +neuman-neural-human-radiance-field-from-a-single-video,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920400.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920400-supp.zip +tava-template-free-animatable-volumetric-actors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920417.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920417-supp.pdf +easnet-searching-elastic-and-accurate-network-architecture-for-stereo-matching,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920434.pdf, +relative-pose-from-sift-features,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920451.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920451-supp.zip +selection-and-cross-similarity-for-event-image-deep-stereo,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920467.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920467-supp.pdf +d3net-a-unified-speaker-listener-architecture-for-3d-dense-captioning-and-visual-grounding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920484.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920484-supp.pdf +circle-convolutional-implicit-reconstruction-and-completion-for-large-scale-indoor-scene,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920502.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920502-supp.pdf +particlesfm-exploiting-dense-point-trajectories-for-localizing-moving-cameras-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920519.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920519-supp.pdf +4dcontrast-contrastive-learning-with-dynamic-correspondences-for-3d-scene-understanding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920539.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920539-supp.pdf +few-zero-level-set-shot-learning-of-shape-signed-distance-functions-in-feature-space,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920556.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920556-supp.pdf +solution-space-analysis-of-essential-matrix-based-on-algebraic-error-minimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920574.pdf, +approximate-differentiable-rendering-with-algebraic-surfaces,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920591.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920591-supp.pdf +covispose-co-visibility-pose-transformer-for-wide-baseline-relative-pose-estimation-in-360deg-indoor-panoramas,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920610.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920610-supp.pdf +affine-correspondences-between-multi-camera-systems-for-6dof-relative-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920629.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920629-supp.zip +graphfit-learning-multi-scale-graph-convolutional-representation-for-point-cloud-normal-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920646.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920646-supp.pdf +is-mvsnet-importance-sampling-based-mvsnet,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920663.pdf, +point-scene-understanding-via-disentangled-instance-mesh-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920679.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920679-supp.pdf +diffustereo-high-quality-human-reconstruction-via-diffusion-based-stereo-using-sparse-cameras,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920697.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920697-supp.pdf +space-partitioning-ransac,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920715.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136920715-supp.zip +simplerecon-3d-reconstruction-without-3d-convolutions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930001-supp.pdf +structure-and-motion-from-casual-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930020.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930020-supp.pdf +what-matters-for-3d-scene-flow-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930036.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930036-supp.pdf +correspondence-reweighted-translation-averaging,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930053.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930053-supp.pdf +neural-strands-learning-hair-geometry-and-appearance-from-multi-view-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930070-supp.zip +graphcspn-geometry-aware-depth-completion-via-dynamic-gcns,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930087.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930087-supp.zip +objects-can-move-3d-change-detection-by-geometric-transformation-consistency,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930104.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930104-supp.pdf +language-grounded-indoor-3d-semantic-segmentation-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930121.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930121-supp.zip +beyond-periodicity-towards-a-unifying-framework-for-activations-in-coordinate-mlps,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930139.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930139-supp.pdf +deforming-radiance-fields-with-cages,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930155.pdf, +flex-extrinsic-parameters-free-multi-view-3d-human-motion-reconstruction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930172.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930172-supp.pdf +mode-multi-view-omnidirectional-depth-estimation-with-360deg-cameras,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930192.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930192-supp.pdf +gigadepth-learning-depth-from-structured-light-with-branching-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930209.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930209-supp.pdf +activenerf-learning-where-to-see-with-uncertainty-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930225.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930225-supp.pdf +posernet-refining-relative-camera-poses-exploiting-object-detections,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930242.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930242-supp.pdf +gaussian-activated-neural-radiance-fields-for-high-fidelity-reconstruction-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930259.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930259-supp.pdf +unbiased-gradient-estimation-for-differentiable-surface-splatting-via-poisson-sampling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930276.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930276-supp.pdf +towards-learning-neural-representations-from-shadows,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930295.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930295-supp.pdf +class-incremental-novel-class-discovery,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930312.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930312-supp.pdf +unknown-oriented-learning-for-open-set-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930328.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930328-supp.pdf +prototype-guided-continual-adaptation-for-class-incremental-unsupervised-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930345.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930345-supp.pdf +decouplenet-decoupled-network-for-domain-adaptive-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930362.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930362-supp.pdf +class-agnostic-object-counting-robust-to-intraclass-diversity,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930380.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930380-supp.pdf +burn-after-reading-online-adaptation-for-cross-domain-streaming-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930396.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930396-supp.pdf +mind-the-gap-in-distilling-stylegans,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930416.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930416-supp.pdf +improving-test-time-adaptation-via-shift-agnostic-weight-regularization-and-nearest-source-prototypes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930433.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930433-supp.pdf +learning-instance-specific-adaptation-for-cross-domain-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930451.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930451-supp.pdf +regioncl-exploring-contrastive-region-pairs-for-self-supervised-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930468.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930468-supp.pdf +long-tailed-class-incremental-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930486.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930486-supp.pdf +dlcft-deep-linear-continual-fine-tuning-for-general-incremental-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930503.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930503-supp.pdf +adversarial-partial-domain-adaptation-by-cycle-inconsistency,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930520.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930520-supp.pdf +combating-label-distribution-shift-for-active-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930539.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930539-supp.pdf +gipso-geometrically-informed-propagation-for-online-adaptation-in-3d-lidar-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930557.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930557-supp.pdf +cosmix-compositional-semantic-mix-for-domain-adaptation-in-3d-lidar-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930575.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930575-supp.pdf +a-unified-framework-for-domain-adaptive-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930592.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930592-supp.pdf +a-broad-study-of-pre-training-for-domain-generalization-and-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930609.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930609-supp.pdf +prior-knowledge-guided-unsupervised-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930628.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930628-supp.pdf +gcisg-guided-causal-invariant-learning-for-improved-syn-to-real-generalization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930644.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930644-supp.pdf +acrofod-an-adaptive-method-for-cross-domain-few-shot-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930661.pdf, +unsupervised-domain-adaptation-for-one-stage-object-detector-using-offsets-to-bounding-box,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930679.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930679-supp.pdf +visual-prompt-tuning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930696.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930696-supp.pdf +quasi-balanced-self-training-on-noise-aware-synthesis-of-object-point-clouds-for-closing-domain-gap,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930715.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136930715-supp.pdf +interpretable-open-set-domain-adaptation-via-angular-margin-separation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940001-supp.pdf +tacs-taxonomy-adaptive-cross-domain-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940019-supp.pdf +prototypical-contrast-adaptation-for-domain-adaptive-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940036.pdf, +rbc-rectifying-the-biased-context-in-continual-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940054.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940054-supp.pdf +factorizing-knowledge-in-neural-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940072.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940072-supp.pdf +contrastive-vicinal-space-for-unsupervised-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940090.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940090-supp.pdf +cross-modal-knowledge-transfer-without-task-relevant-source-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940108.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940108-supp.pdf +online-domain-adaptation-for-semantic-segmentation-in-ever-changing-conditions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940125.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940125-supp.pdf +source-free-video-domain-adaptation-by-learning-temporal-consistency-for-action-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940144.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940144-supp.pdf +bmd-a-general-class-balanced-multicentric-dynamic-prototype-strategy-for-source-free-domain-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940161.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940161-supp.pdf +generalized-brain-image-synthesis-with-transferable-convolutional-sparse-coding-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940178.pdf, +incomplete-multi-view-domain-adaptation-via-channel-enhancement-and-knowledge-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940194.pdf, +distpro-searching-a-fast-knowledge-distillation-process-via-meta-optimization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940211.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940211-supp.pdf +ml-bpm-multi-teacher-learning-with-bidirectional-photometric-mixing-for-open-compound-domain-adaptation-in-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940228.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940228-supp.pdf +pactran-pac-bayesian-metrics-for-estimating-the-transferability-of-pretrained-models-to-classification-tasks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940244.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940244-supp.pdf +personalized-education-blind-knowledge-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940262.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940262-supp.pdf +not-all-models-are-equal-predicting-model-transferability-in-a-self-challenging-fisher-space,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940279.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940279-supp.pdf +how-stable-are-transferability-metrics-evaluations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940296.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940296-supp.pdf +attention-diversification-for-domain-generalization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940315.pdf, +ess-learning-event-based-semantic-segmentation-from-still-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940334.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940334-supp.pdf +an-efficient-spatio-temporal-pyramid-transformer-for-action-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940350.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940350-supp.pdf +human-trajectory-prediction-via-neural-social-physics,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940368.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940368-supp.pdf +towards-open-set-video-anomaly-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940387.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940387-supp.pdf +eclipse-efficient-long-range-video-retrieval-using-sight-and-sound,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940405.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940405-supp.zip +joint-modal-label-denoising-for-weakly-supervised-audio-visual-video-parsing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940424.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940424-supp.pdf +less-than-few-self-shot-video-instance-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940442.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940442-supp.pdf +adaptive-face-forgery-detection-in-cross-domain,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940460.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940460-supp.pdf +real-time-online-video-detection-with-temporal-smoothing-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940478.pdf, +tallformer-temporal-action-localization-with-a-long-memory-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940495.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940495-supp.pdf +mining-relations-among-cross-frame-affinities-for-video-semantic-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940513.pdf, +tl-dw-summarizing-instructional-videos-with-task-relevance-cross-modal-saliency,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940530.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940530-supp.pdf +rethinking-learning-approaches-for-long-term-action-anticipation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940547.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940547-supp.zip +dualformer-local-global-stratified-transformer-for-efficient-video-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940566.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940566-supp.pdf +hierarchical-feature-alignment-network-for-unsupervised-video-object-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940584.pdf, +pac-net-highlight-your-video-via-history-preference-modeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940602.pdf, +how-severe-is-benchmark-sensitivity-in-video-self-supervised-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940620.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940620-supp.pdf +a-sliding-window-scheme-for-online-temporal-action-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940640.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940640-supp.pdf +era-expert-retrieval-and-assembly-for-early-action-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940657.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940657-supp.pdf +dual-perspective-network-for-audio-visual-event-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940676.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940676-supp.pdf +nsnet-non-saliency-suppression-sampler-for-efficient-video-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940692.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940692-supp.pdf +video-activity-localisation-with-uncertainties-in-temporal-boundary,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940710.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940710-supp.pdf +temporal-saliency-query-network-for-efficient-video-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940727.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136940727-supp.pdf +efficient-one-stage-video-object-detection-by-exploiting-temporal-consistency,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950001-supp.pdf +leveraging-action-affinity-and-continuity-for-semi-supervised-temporal-action-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950017.pdf, +spotting-temporally-precise-fine-grained-events-in-video,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950033.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950033-supp.pdf +unified-fully-and-timestamp-supervised-temporal-action-segmentation-via-sequence-to-sequence-translation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950052.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950052-supp.pdf +efficient-video-transformers-with-spatial-temporal-token-selection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950068.pdf, +long-movie-clip-classification-with-state-space-video-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950086.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950086-supp.pdf +prompting-visual-language-models-for-efficient-video-understanding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950104.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950104-supp.zip +asymmetric-relation-consistency-reasoning-for-video-relation-grounding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950124.pdf, +self-supervised-social-relation-representation-for-human-group-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950140.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950140-supp.pdf +k-centered-patch-sampling-for-efficient-video-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950157.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950157-supp.pdf +a-deep-moving-camera-background-model,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950175.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950175-supp.zip +graphvid-it-only-takes-a-few-nodes-to-understand-a-video,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950192.pdf, +delta-distillation-for-efficient-video-processing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950209.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950209-supp.pdf +morphmlp-an-efficient-mlp-like-backbone-for-spatial-temporal-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950226.pdf, +composer-compositional-reasoning-of-group-activity-in-videos-with-keypoint-only-modality,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950245.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950245-supp.pdf +e-nerv-expedite-neural-video-representation-with-disentangled-spatial-temporal-context,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950263.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950263-supp.pdf +tdvit-temporal-dilated-video-transformer-for-dense-video-tasks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950281.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950281-supp.pdf +semi-supervised-learning-of-optical-flow-by-flow-supervisor,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950298.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950298-supp.pdf +flow-graph-to-video-grounding-for-weakly-supervised-multi-step-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950315.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950315-supp.pdf +deep-360deg-optical-flow-estimation-based-on-multi-projection-fusion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950332.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950332-supp.zip +maclr-motion-aware-contrastive-learning-of-representations-for-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950349.pdf, +learning-long-term-spatial-temporal-graphs-for-active-speaker-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950367.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950367-supp.zip +frozen-clip-models-are-efficient-video-learners,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950384.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950384-supp.pdf +pip-physical-interaction-prediction-via-mental-simulation-with-span-selection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950401.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950401-supp.pdf +panoramic-vision-transformer-for-saliency-detection-in-360deg-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950419.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950419-supp.pdf +bayesian-tracking-of-video-graphs-using-joint-kalman-smoothing-and-registration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950436.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950436-supp.zip +motion-sensitive-contrastive-learning-for-self-supervised-video-representation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950453.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950453-supp.pdf +dynamic-temporal-filtering-in-video-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950470.pdf, +tip-adapter-training-free-adaption-of-clip-for-few-shot-classification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950487.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950487-supp.pdf +temporal-lift-pooling-for-continuous-sign-language-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950506.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950506-supp.pdf +more-multi-order-relation-mining-for-dense-captioning-in-3d-scenes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950523.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950523-supp.pdf +siri-a-simple-selective-retraining-mechanism-for-transformer-based-visual-grounding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950541.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950541-supp.pdf +cross-modal-prototype-driven-network-for-radiology-report-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950558.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950558-supp.pdf +tm2t-stochastic-and-tokenized-modeling-for-the-reciprocal-generation-of-3d-human-motions-and-texts,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950575.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950575-supp.pdf +seqtr-a-simple-yet-universal-network-for-visual-grounding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950593.pdf, +vtc-improving-video-text-retrieval-with-user-comments,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950611.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950611-supp.pdf +fashionvil-fashion-focused-vision-and-language-representation-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950629.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950629-supp.pdf +weakly-supervised-grounding-for-vqa-in-vision-language-transformers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950647.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950647-supp.pdf +automatic-dense-annotation-of-large-vocabulary-sign-language-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950666.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950666-supp.pdf +miles-visual-bert-pre-training-with-injected-language-semantics-for-video-text-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950685.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950685-supp.pdf +geb-a-benchmark-for-generic-event-boundary-captioning-grounding-and-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950703.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950703-supp.pdf +a-simple-and-robust-correlation-filtering-method-for-text-based-person-search,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136950719.pdf, +making-the-most-of-text-semantics-to-improve-biomedical-vision-language-processing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960001-supp.pdf +generative-negative-text-replay-for-continual-vision-language-pretraining,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960022.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960022-supp.pdf +video-graph-transformer-for-video-question-answering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960039.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960039-supp.pdf +trace-controlled-text-to-image-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960058.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960058-supp.pdf +video-question-answering-with-iterative-video-text-co-tokenization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960075.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960075-supp.pdf +rethinking-data-augmentation-for-robust-visual-question-answering,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960094.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960094-supp.pdf +explicit-image-caption-editing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960111.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960111-supp.pdf +can-shuffling-video-benefit-temporal-bias-problem-a-novel-training-framework-for-temporal-grounding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960128.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960128-supp.pdf +reliable-visual-question-answering-abstain-rather-than-answer-incorrectly,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960146.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960146-supp.pdf +grit-faster-and-better-image-captioning-transformer-using-dual-visual-features,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960165.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960165-supp.pdf +selective-query-guided-debiasing-for-video-corpus-moment-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960183.pdf, +spatial-and-visual-perspective-taking-via-view-rotation-and-relation-reasoning-for-embodied-reference-understanding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960199.pdf, +object-centric-unsupervised-image-captioning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960217.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960217-supp.pdf +contrastive-vision-language-pre-training-with-limited-resources,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960234.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960234-supp.pdf +learning-linguistic-association-towards-efficient-text-video-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960251.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960251-supp.pdf +assister-assistive-navigation-via-conditional-instruction-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960269.pdf, +x-detr-a-versatile-architecture-for-instance-wise-vision-language-tasks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960288.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960288-supp.pdf +learning-disentanglement-with-decoupled-labels-for-vision-language-navigation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960305.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960305-supp.pdf +switch-bert-learning-to-model-multimodal-interactions-by-switching-attention-and-input,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960325.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960325-supp.pdf +word-level-fine-grained-story-visualization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960342.pdf, +unifying-event-detection-and-captioning-as-sequence-generation-via-pre-training,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960358.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960358-supp.pdf +multimodal-transformer-with-variable-length-memory-for-vision-and-language-navigation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960375.pdf, +fine-grained-visual-entailment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960393.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960393-supp.pdf +bottom-up-top-down-detection-transformers-for-language-grounding-in-images-and-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960411.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960411-supp.pdf +new-datasets-and-models-for-contextual-reasoning-in-visual-dialog,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960428.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960428-supp.pdf +visagesyntalk-unseen-speaker-video-to-speech-synthesis-via-speech-visage-feature-selection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960445.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960445-supp.zip +classification-regression-for-chart-comprehension,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960462.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960462-supp.pdf +assistq-affordance-centric-question-driven-task-completion-for-egocentric-assistant,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960478.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960478-supp.pdf +findit-generalized-localization-with-natural-language-queries,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960495.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960495-supp.pdf +unitab-unifying-text-and-box-outputs-for-grounded-vision-language-modeling,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960514.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960514-supp.pdf +scaling-open-vocabulary-image-segmentation-with-image-level-labels,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960532.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960532-supp.pdf +the-abduction-of-sherlock-holmes-a-dataset-for-visual-abductive-reasoning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960549.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960549-supp.pdf +speaker-adaptive-lip-reading-with-user-dependent-padding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960567.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960567-supp.pdf +tise-bag-of-metrics-for-text-to-image-synthesis-evaluation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960585.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960585-supp.pdf +semaug-semantically-meaningful-image-augmentations-for-object-detection-through-language-grounding,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960602.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960602-supp.pdf +referring-object-manipulation-of-natural-images-with-conditional-classifier-free-guidance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960619.pdf, +newsstories-illustrating-articles-with-visual-summaries,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960636.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960636-supp.pdf +webly-supervised-concept-expansion-for-general-purpose-vision-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960654.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960654-supp.pdf +fedvln-privacy-preserving-federated-vision-and-language-navigation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960673.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960673-supp.pdf +coder-coupled-diversity-sensitive-momentum-contrastive-learning-for-image-text-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960691.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960691-supp.pdf +language-driven-artistic-style-transfer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960708.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960708-supp.pdf +single-stream-multi-level-alignment-for-vision-language-pretraining,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960725.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136960725-supp.pdf +most-and-least-retrievable-images-in-visual-language-query-systems,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970001-supp.pdf +sports-video-analysis-on-large-scale-data,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970019.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970019-supp.pdf +grounding-visual-representations-with-texts-for-domain-generalization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970037.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970037-supp.pdf +bridging-the-visual-semantic-gap-in-vln-via-semantically-richer-instructions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970054.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970054-supp.pdf +storydall-e-adapting-pretrained-text-to-image-transformers-for-story-continuation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970070-supp.pdf +vqgan-clip-open-domain-image-generation-and-editing-with-natural-language-guidance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970088.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970088-supp.pdf +semantic-aware-implicit-neural-audio-driven-video-portrait-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970105.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970105-supp.pdf +end-to-end-active-speaker-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970124.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970124-supp.pdf +emotion-recognition-for-multiple-context-awareness,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970141.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970141-supp.pdf +adaptive-fine-grained-sketch-based-image-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970160.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970160-supp.pdf +quantized-gan-for-complex-music-generation-from-dance-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970177.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970177-supp.pdf +uncertainty-aware-multi-modal-learning-via-cross-modal-random-network-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970195.pdf, +localizing-visual-sounds-the-easy-way,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970212.pdf, +learning-visual-styles-from-audio-visual-associations,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970229.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970229-supp.pdf +remote-respiration-monitoring-of-moving-person-using-radio-signals,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970248.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970248-supp.pdf +camera-pose-estimation-and-localization-with-active-audio-sensing,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970266.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970266-supp.pdf +pacs-a-dataset-for-physical-audiovisual-commonsense-reasoning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970286.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970286-supp.zip +vovit-low-latency-graph-based-audio-visual-voice-separation-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970304.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970304-supp.zip +telepresence-video-quality-assessment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970321.pdf, +multimae-multi-modal-multi-task-masked-autoencoders,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970341.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970341-supp.zip +audioscopev2-audio-visual-attention-architectures-for-calibrated-open-domain-on-screen-sound-separation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970360.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970360-supp.pdf +audio-visual-segmentation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970378.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970378-supp.pdf +unsupervised-night-image-enhancement-when-layer-decomposition-meets-light-effects-suppression,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970396.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970396-supp.pdf +relationformer-a-unified-framework-for-image-to-graph-generation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970414.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970414-supp.pdf +gama-cross-view-video-geo-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970432.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970432-supp.pdf +revisiting-a-knn-based-image-classification-system-with-high-capacity-storage,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970449.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970449-supp.pdf +geometric-representation-learning-for-document-image-rectification,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970466.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970466-supp.pdf +s2-ver-semi-supervised-visual-emotion-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970483.pdf, +image-coding-for-machines-with-omnipotent-feature-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970500.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970500-supp.pdf +feature-representation-learning-for-unsupervised-cross-domain-image-retrieval,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970518.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970518-supp.pdf +fashionformer-a-simple-effective-and-unified-baseline-for-human-fashion-segmentation-and-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970534.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970534-supp.pdf +semantic-guided-multi-mask-image-harmonization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970552.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970552-supp.pdf +learning-an-isometric-surface-parameterization-for-texture-unwrapping,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970568.pdf, +towards-regression-free-neural-networks-for-diverse-compute-platforms,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970587.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970587-supp.pdf +relationship-spatialization-for-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970603.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970603-supp.pdf +image2point-3d-point-cloud-understanding-with-2d-image-pretrained-models,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970625.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970625-supp.pdf +far-fourier-aerial-video-recognition,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970644.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970644-supp.zip +translating-a-visual-lego-manual-to-a-machine-executable-plan,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970663.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970663-supp.pdf +fabric-material-recovery-from-video-using-multi-scale-geometric-auto-encoder,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970680.pdf, +megba-a-gpu-based-distributed-library-for-large-scale-bundle-adjustment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970698.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970698-supp.pdf +the-one-where-they-reconstructed-3d-humans-and-environments-in-tv-shows,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970714.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136970714-supp.pdf +talisman-targeted-active-learning-for-object-detection-with-rare-classes-and-slices-using-submodular-mutual-information,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980001.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980001-supp.pdf +an-efficient-person-clustering-algorithm-for-open-checkout-free-groceries,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980017.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980017-supp.zip +pop-mining-potential-performance-of-new-fashion-products-via-webly-cross-modal-query-expansion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980034.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980034-supp.pdf +pose-forecasting-in-industrial-human-robot-collaboration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980051.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980051-supp.pdf +actor-centered-representations-for-action-localization-in-streaming-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980070-supp.zip +bandwidth-aware-adaptive-codec-for-dnn-inference-offloading-in-iot,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980087.pdf, +domain-knowledge-informed-self-supervised-representations-for-workout-form-assessment,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980104.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980104-supp.zip +responsive-listening-head-generation-a-benchmark-dataset-and-baseline,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980122.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980122-supp.pdf +towards-scale-aware-robust-and-generalizable-unsupervised-monocular-depth-estimation-by-integrating-imu-motion-dynamics,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980140.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980140-supp.pdf +tips-text-induced-pose-synthesis,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980157.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980157-supp.pdf +addressing-heterogeneity-in-federated-learning-via-distributional-transformation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980175.pdf, +where-in-the-world-is-this-image-transformer-based-geo-localization-in-the-wild,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980193.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980193-supp.pdf +colorization-for-in-situ-marine-plankton-images,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980212.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980212-supp.pdf +efficient-deep-visual-and-inertial-odometry-with-adaptive-visual-modality-selection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980229.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980229-supp.pdf +a-sketch-is-worth-a-thousand-words-image-retrieval-with-text-and-sketch,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980247.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980247-supp.pdf +a-cloud-3d-dataset-and-application-specific-learned-image-compression-in-cloud-3d,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980265.pdf, +autotransition-learning-to-recommend-video-transition-effects,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980282.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980282-supp.zip +online-segmentation-of-lidar-sequences-dataset-and-algorithm,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980298.pdf, +open-world-semantic-segmentation-for-lidar-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980315.pdf, +king-generating-safety-critical-driving-scenarios-for-robust-imitation-via-kinematics-gradients,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980332.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980332-supp.pdf +differentiable-raycasting-for-self-supervised-occupancy-forecasting,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980349.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980349-supp.zip +inaction-interpretable-action-decision-making-for-autonomous-driving,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980365.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980365-supp.pdf +cramnet-camera-radar-fusion-with-ray-constrained-cross-attention-for-robust-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980382.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980382-supp.pdf +coda-a-real-world-road-corner-case-dataset-for-object-detection-in-autonomous-driving,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980399.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980399-supp.pdf +motion-inspired-unsupervised-perception-and-prediction-in-autonomous-driving,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980416.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980416-supp.pdf +stretchbev-stretching-future-instance-prediction-spatially-and-temporally,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980436.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980436-supp.pdf +rclane-relay-chain-prediction-for-lane-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980453.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980453-supp.pdf +drive-segment-unsupervised-semantic-segmentation-of-urban-scenes-via-cross-modal-distillation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980469.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980469-supp.pdf +centerformer-center-based-transformer-for-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980487.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980487-supp.pdf +physical-attack-on-monocular-depth-estimation-with-optimal-adversarial-patches,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980504.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980504-supp.pdf +st-p3-end-to-end-vision-based-autonomous-driving-via-spatial-temporal-feature-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980522.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980522-supp.pdf +persformer-3d-lane-detection-via-perspective-transformer-and-the-openlane-benchmark,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980539.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980539-supp.pdf +pointfix-learning-to-fix-domain-bias-for-robust-online-stereo-adaptation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980557.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980557-supp.zip +brnet-exploring-comprehensive-features-for-monocular-depth-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980574.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980574-supp.pdf +siamdoge-domain-generalizable-semantic-segmentation-using-siamese-network,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980590.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980590-supp.pdf +context-aware-streaming-perception-in-dynamic-environments,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980608.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980608-supp.zip +spot-spatiotemporal-modeling-for-3d-object-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980624.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980624-supp.pdf +multimodal-transformer-for-automatic-3d-annotation-and-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980641.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980641-supp.pdf +dynamic-3d-scene-analysis-by-point-cloud-accumulation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980658.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980658-supp.pdf +homogeneous-multi-modal-feature-fusion-and-interaction-for-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980675.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980675-supp.pdf +jperceiver-joint-perception-network-for-depth-pose-and-layout-estimation-in-driving-scenes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980692.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980692-supp.pdf +semi-supervised-3d-object-detection-with-proficient-teachers,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980710.pdf, +point-cloud-compression-with-sibling-context-and-surface-priors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980726.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136980726-supp.pdf +lane-detection-transformer-based-on-multi-frame-horizontal-and-vertical-attention-and-visual-transformer-module,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990001.pdf, +proposalcontrast-unsupervised-pre-training-for-lidar-based-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990017.pdf, +pretram-self-supervised-pre-training-via-connecting-trajectory-and-map,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990034.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990034-supp.pdf +master-of-all-simultaneous-generalization-of-urban-scene-segmentation-to-all-adverse-weather-conditions,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990051.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990051-supp.pdf +less-label-efficient-semantic-segmentation-for-lidar-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990070.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990070-supp.pdf +visual-cross-view-metric-localization-with-dense-uncertainty-estimates,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990089.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990089-supp.zip +v2x-vit-vehicle-to-everything-cooperative-perception-with-vision-transformer,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990106.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990106-supp.pdf +devnet-self-supervised-monocular-depth-learning-via-density-volume-construction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990123.pdf, +action-based-contrastive-learning-for-trajectory-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990140.pdf, +radatron-accurate-detection-using-multi-resolution-cascaded-mimo-radar,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990157.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990157-supp.zip +lidar-distillation-bridging-the-beam-induced-domain-gap-for-3d-object-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990175.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990175-supp.zip +efficient-point-cloud-segmentation-with-geometry-aware-sparse-networks,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990193.pdf, +fh-net-a-fast-hierarchical-network-for-scene-flow-estimation-on-real-world-point-clouds,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990210.pdf, +spatialdetr-robust-scalable-transformer-based-3d-object-detection-from-multi-view-camera-images-with-global-cross-sensor-attention,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990226.pdf, +pixel-wise-energy-biased-abstention-learning-for-anomaly-segmentation-on-complex-urban-driving-scenes,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990242.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990242-supp.pdf +rethinking-closed-loop-training-for-autonomous-driving,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990259.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990259-supp.zip +slide-self-supervised-lidar-de-snowing-through-reconstruction-difficulty,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990277.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990277-supp.pdf +generative-meta-adversarial-network-for-unseen-object-navigation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990295.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990295-supp.pdf +object-manipulation-via-visual-target-localization,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990314.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990314-supp.zip +moda-map-style-transfer-for-self-supervised-domain-adaptation-of-embodied-agents,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990332.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990332-supp.zip +housekeep-tidying-virtual-households-using-commonsense-reasoning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990350.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990350-supp.pdf +domain-randomization-enhanced-depth-simulation-and-restoration-for-perceiving-and-grasping-specular-and-transparent-objects,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990369.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990369-supp.pdf +resolving-copycat-problems-in-visual-imitation-learning-via-residual-action-prediction,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990386.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990386-supp.pdf +opd-single-view-3d-openable-part-detection,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990404.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990404-supp.zip +airdet-few-shot-detection-without-fine-tuning-for-autonomous-exploration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990421.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990421-supp.pdf +transgrasp-grasp-pose-estimation-of-a-category-of-objects-by-transferring-grasps-from-only-one-labeled-instance,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990438.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990438-supp.pdf +starformer-transformer-with-state-action-reward-representations-for-visual-reinforcement-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990455.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990455-supp.pdf +tidee-tidying-up-novel-rooms-using-visuo-semantic-commonsense-priors,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990473.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990473-supp.pdf +learning-efficient-multi-agent-cooperative-visual-exploration,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990491.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990491-supp.pdf +zero-shot-category-level-object-pose-estimation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990509.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990509-supp.pdf +sim-to-real-6d-object-pose-estimation-via-iterative-self-training-for-robotic-bin-picking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990526.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990526-supp.pdf +active-audio-visual-separation-of-dynamic-sound-sources,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990543.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990543-supp.pdf +dexmv-imitation-learning-for-dexterous-manipulation-from-human-videos,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990562.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990562-supp.pdf +sim-2-sim-transfer-for-vision-and-language-navigation-in-continuous-environments,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990580.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990580-supp.zip +style-agnostic-reinforcement-learning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990596.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990596-supp.zip +self-supervised-interactive-object-segmentation-through-a-singulation-and-grasping-approach,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990613.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990613-supp.pdf +learning-from-unlabeled-3d-environments-for-vision-and-language-navigation,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990630.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990630-supp.pdf +bodyslam-joint-camera-localisation-mapping-and-human-motion-tracking,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990648.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990648-supp.zip +fusionvae-a-deep-hierarchical-variational-autoencoder-for-rgb-image-fusion,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990666.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990666-supp.pdf +learning-algebraic-representation-for-systematic-generalization-in-abstract-reasoning,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990683.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990683-supp.pdf +video-dialog-as-conversation-about-objects-living-in-space-time,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990701.pdf,https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990701-supp.pdf diff --git a/te_u/paper_down_load/eccv_download.py b/te_u/paper_down_load/eccv_download.py new file mode 100644 index 0000000..29412ae --- /dev/null +++ b/te_u/paper_down_load/eccv_download.py @@ -0,0 +1,658 @@ +from bs4 import BeautifulSoup +import pickle +import os + +os.environ['http_proxy'] = '127.0.0.1:7890' +os.environ['https_proxy'] = '127.0.0.1:7890' + +from tqdm import tqdm +from slugify import slugify +import csv +import sys + +import urllib +import random +from urllib.error import URLError, HTTPError + +import requests + + +class Downloader: + def __init__(self, downloader=None, is_random_step=None): + pass + + def download(self, urls=None, save_path=None, time_sleep_in_seconds=None): + print(urls) + headers = { + 'User-Agent': + 'Mozilla/5.0 (Windows NT 6.1; WOW64; rv:23.0) Gecko/20100101 Firefox/23.0'} + content = urlopen_with_retry(url=urls, headers=headers) + with open(save_path, 'wb') as f: + f.write(content) + + +def download_from_csv_i( + postfix=None, save_dir=None, csv_file_path=None, is_download_main_paper=True, + is_download_bib=True, is_download_supplement=True, + time_step_in_seconds=5, total_paper_number=None, + downloader='IDM', is_random_step=True): + """ + download paper, bibtex and supplement files and save them to + save_dir/main_paper and save_dir/supplement respectively + :param postfix: str, postfix that will be added at the end of papers' title + :param save_dir: str, paper and supplement material's save path + :param csv_file_path: str, the full path to csv file + :param is_download_main_paper: bool, True for downloading main paper + :param is_download_supplement: bool, True for downloading supplemental + material + :param time_step_in_seconds: int, the interval time between two downloading + request in seconds + :param total_paper_number: int, the total number of papers that is going to + download + :param downloader: str, the downloader to download, could be 'IDM' or None, + default to 'IDM'. + :param is_random_step: bool, whether random sample the time step between two + adjacent download requests. If True, the time step will be sampled + from Uniform(0.5t, 1.5t), where t is the given time_step_in_seconds. + Default: True. + :return: True + """ + downloader = Downloader( + downloader=downloader, is_random_step=is_random_step) + if not os.path.exists(csv_file_path): + raise ValueError(f'ERROR: file not found in {csv_file_path}!!!') + + main_save_path = os.path.join(save_dir, 'main_paper') + if is_download_main_paper: + os.makedirs(main_save_path, exist_ok=True) + if is_download_supplement: + supplement_save_path = os.path.join(save_dir, 'supplement') + os.makedirs(supplement_save_path, exist_ok=True) + + error_log = [] + with open(csv_file_path, newline='') as csvfile: + myreader = csv.DictReader(csvfile, delimiter=',') + pbar = tqdm(myreader, total=total_paper_number) + i = 0 + for this_paper in pbar: + is_download_bib &= ('bib' in this_paper) + is_grouped = ('group' in this_paper) + i += 1 + # get title + if is_grouped: + group = slugify(this_paper['group']) + title = slugify(this_paper['title']) + if total_paper_number is not None: + pbar.set_description( + f'Downloading {postfix} paper {i} /{total_paper_number}') + else: + pbar.set_description(f'Downloading {postfix} paper {i}') + this_paper_main_path = os.path.join( + main_save_path, f'{title}_{postfix}.pdf') + if is_grouped: + this_paper_main_path = os.path.join( + main_save_path, group, f'{title}_{postfix}.pdf') + if is_download_supplement: + this_paper_supp_path_no_ext = os.path.join( + supplement_save_path, f'{title}_{postfix}_supp.') + if is_grouped: + this_paper_supp_path_no_ext = os.path.join( + supplement_save_path, group, f'{title}_{postfix}_supp.') + if '' != this_paper['supplemental link'] and os.path.exists( + this_paper_main_path) and \ + (os.path.exists( + this_paper_supp_path_no_ext + 'zip') or + os.path.exists( + this_paper_supp_path_no_ext + 'pdf')): + continue + elif '' == this_paper['supplemental link'] and \ + os.path.exists(this_paper_main_path): + continue + elif os.path.exists(this_paper_main_path): + continue + if 'error' == this_paper['main link']: + error_log.append((title, 'no MAIN link')) + elif '' != this_paper['main link']: + if is_grouped: + if is_download_main_paper: + os.makedirs(os.path.join(main_save_path, group), + exist_ok=True) + if is_download_supplement: + os.makedirs(os.path.join(supplement_save_path, group), + exist_ok=True) + if is_download_main_paper: + try: + # download paper with IDM + if not os.path.exists(this_paper_main_path): + downloader.download( + urls=this_paper['main link'].replace( + ' ', '%20'), + save_path=os.path.join( + os.getcwd(), this_paper_main_path), + time_sleep_in_seconds=time_step_in_seconds + ) + except Exception as e: + # error_flag = True + print('Error: ' + title + ' - ' + str(e)) + error_log.append((title, this_paper['main link'], + 'main paper download error', str(e))) + # download supp + if is_download_supplement: + # check whether the supp can be downloaded + if not (os.path.exists( + this_paper_supp_path_no_ext + 'zip') or + os.path.exists( + this_paper_supp_path_no_ext + 'pdf')): + if 'error' == this_paper['supplemental link']: + error_log.append((title, 'no SUPPLEMENTAL link')) + elif '' != this_paper['supplemental link']: + supp_type = \ + this_paper['supplemental link'].split('.')[-1] + try: + downloader.download( + urls=this_paper['supplemental link'], + save_path=os.path.join( + os.getcwd(), + this_paper_supp_path_no_ext + supp_type), + time_sleep_in_seconds=time_step_in_seconds + ) + except Exception as e: + # error_flag = True + print('Error: ' + title + ' - ' + str(e)) + error_log.append((title, this_paper[ + 'supplemental link'], + 'supplement download error', + str(e))) + # download bibtex file + if is_download_bib: + bib_path = this_paper_main_path[:-3] + 'bib' + if not os.path.exists(bib_path): + if 'error' == this_paper['bib']: + error_log.append((title, 'no bibtex link')) + elif '' != this_paper['bib']: + try: + downloader.download( + urls=this_paper['bib'], + save_path=os.path.join(os.getcwd(), + bib_path), + time_sleep_in_seconds=time_step_in_seconds + ) + except Exception as e: + # error_flag = True + print('Error: ' + title + ' - ' + str(e)) + error_log.append((title, this_paper['bib'], + 'bibtex download error', + str(e))) + + # 2. write error log + print('write error log') + return True + + +def get_paper_name_link_from_url(url): + headers = { + 'User-Agent': + 'Mozilla/5.0 (Windows NT 6.1; WOW64; rv:23.0) Gecko/20100101 Firefox/23.0'} + paper_dict = dict() + content = urlopen_with_retry(url=url, headers=headers) + soup = BeautifulSoup(content, 'html5lib') + paper_list_bar = tqdm(soup.find_all(['li'], {'class': 'chapter-item content-type-list__item'})) + for paper in paper_list_bar: + try: + title = slugify(paper.find('div', {'class': 'content-type-list__title'}).text) + link = urllib.parse.urljoin(url, paper.find('div', {'class': 'content-type-list__action'}).a.get('href')) + paper_dict[title] = link + except Exception as e: + print(f'ERROR: {str(e)}') + return paper_dict + + +def urlopen_with_retry(url, headers=dict(), retry_time=3, time_out=20, + raise_error_if_failed=True): + """ + load content from url with given headers. Retry if error occurs. + Args: + url (str): url. + headers (dict): request headers. Default: {}. + retry_time (int): max retry time. Default: 3. + time_out (int): time out in seconds. Default: 10. + raise_error_if_failed (bool): whether to raise error if failed. + Default: True. + + Returns: + content(str|None): url content. None will be returned if failed. + + """ + res = requests.get(url=url, headers=headers) + + # req = urllib.request.Request(url=url, headers=headers) + for r in range(retry_time): + try: + # content = urllib.request.urlopen(req, timeout=time_out).read() + content = res.content + return content + except HTTPError as e: + print('The server couldn\'t fulfill the request.') + print('Error code: ', e.code) + s = random.randint(3, 7) + print(f'random sleeping {s} seconds and doing {r + 1}/{retry_time}' + f'-th retrying...') + except URLError as e: + print('We failed to reach a server.') + print('Reason: ', e.reason) + s = random.randint(3, 7) + print(f'random sleeping {s} seconds and doing {r + 1}/{retry_time}' + f'-th retrying...') + if raise_error_if_failed: + raise ValueError(f'Failed to open {url} after trying {retry_time} ' + f'times!') + else: + return None + + +def save_csv(year): + """ + write ECCV papers' and supplemental material's urls in one csv file + :param year: int + :return: True + """ + project_root_folder = r"D:\py\keyan_qingbao\te_u\paper_down_load" + csv_file_pathname = os.path.join( + project_root_folder, 'csv', f'ECCV_{year}.csv') + with open(csv_file_pathname, 'w', newline='') as csvfile: + fieldnames = ['title', 'main link', 'supplemental link'] + writer = csv.DictWriter(csvfile, fieldnames=fieldnames) + writer.writeheader() + headers = { + 'User-Agent': + 'Mozilla/5.0 (Windows NT 6.1; WOW64; rv:23.0) ' + 'Gecko/20100101 Firefox/23.0'} + dat_file_pathname = os.path.join( + project_root_folder, 'urls', f'init_url_ECCV_{year}.dat') + if year >= 2018: + init_url = f'https://www.ecva.net/papers.php' + if os.path.exists(dat_file_pathname): + with open(dat_file_pathname, 'rb') as f: + content = pickle.load(f) + else: + content = urlopen_with_retry(url=init_url, headers=headers) + with open(dat_file_pathname, 'wb') as f: + pickle.dump(content, f) + soup = BeautifulSoup(content, 'html5lib') + paper_list_bar = tqdm(soup.find_all(['dt', 'dd'])) + paper_index = 0 + paper_dict = {'title': '', + 'main link': '', + 'supplemental link': ''} + for paper in paper_list_bar: + is_new_paper = False + + # get title + try: + if 'dt' == paper.name and \ + 'ptitle' == paper.get('class')[0] and \ + year == int(paper.a.get('href').split('_')[1][:4]): # title: + # this_year = int(paper.a.get('href').split('_')[1][:4]) + title = slugify(paper.text.strip()) + paper_dict['title'] = title + paper_index += 1 + paper_list_bar.set_description_str( + f'Downloading paper {paper_index}: {title}') + elif '' != paper_dict['title'] and 'dd' == paper.name: + all_as = paper.find_all('a') + for a in all_as: + if 'pdf' == slugify(a.text.strip()): + main_link = urllib.parse.urljoin(init_url, + a.get('href')) + paper_dict['main link'] = main_link + is_new_paper = True + elif 'supp' == slugify(a.text.strip())[:4]: + supp_link = urllib.parse.urljoin(init_url, + a.get('href')) + paper_dict['supplemental link'] = supp_link + break + except: + pass + if is_new_paper: + writer.writerow(paper_dict) + paper_dict = {'title': '', + 'main link': '', + 'supplemental link': ''} + else: + init_url = f'http://www.eccv{year}.org/main-conference/' + if os.path.exists(dat_file_pathname): + with open(dat_file_pathname, 'rb') as f: + content = pickle.load(f) + else: + content = urlopen_with_retry(url=init_url, headers=headers) + with open(dat_file_pathname, 'wb') as f: + pickle.dump(content, f) + soup = BeautifulSoup(content, 'html5lib') + paper_list_bar = tqdm( + soup.find('div', {'class': 'entry-content'}).find_all(['p'])) + paper_index = 0 + paper_dict = {'title': '', + 'main link': '', + 'supplemental link': ''} + for paper in paper_list_bar: + try: + if len(paper.find_all(['strong'])) and len( + paper.find_all(['a'])) and len(paper.find_all(['img'])): + paper_index += 1 + title = slugify(paper.find('strong').text) + paper_dict['title'] = title + paper_list_bar.set_description_str( + f'Downloading paper {paper_index}: {title}') + main_link = paper.find('a').get('href') + paper_dict['main link'] = main_link + writer.writerow(paper_dict) + paper_dict = {'title': '', + 'main link': '', + 'supplemental link': ''} + except Exception as e: + print(f'ERROR: {str(e)}') + return paper_index + + +def download_from_csv( + year, save_dir, is_download_supplement=True, time_step_in_seconds=5, + total_paper_number=None, + is_workshops=False, downloader='IDM'): + """ + download all ECCV paper and supplement files given year, restore in + save_dir/main_paper and save_dir/supplement respectively + :param year: int, ECCV year, such 2019 + :param save_dir: str, paper and supplement material's save path + :param is_download_supplement: bool, True for downloading supplemental + material + :param time_step_in_seconds: int, the interval time between two downlaod + request in seconds + :param total_paper_number: int, the total number of papers that is going + to download + :param is_workshops: bool, is to download workshops from csv file. + :param downloader: str, the downloader to download, could be 'IDM' or + 'Thunder', default to 'IDM' + :return: True + """ + postfix = f'ECCV_{year}' + if is_workshops: + postfix = f'ECCV_WS_{year}' + csv_file_name = f'ECCV_{year}.csv' if not is_workshops else \ + f'ECCV_WS_{year}.csv' + project_root_folder = r"D:\py\keyan_qingbao\te_u\paper_down_load" + csv_file_name = os.path.join(project_root_folder, 'csv', csv_file_name) + download_from_csv_i( + postfix=postfix, + save_dir=save_dir, + csv_file_path=csv_file_name, + is_download_supplement=is_download_supplement, + time_step_in_seconds=time_step_in_seconds, + total_paper_number=total_paper_number, + downloader=downloader + ) + + +def download_from_springer( + year, save_dir, is_workshops=False, time_sleep_in_seconds=5, + downloader='IDM'): + os.makedirs(save_dir, exist_ok=True) + if 2018 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-030-01246-5', + 'https://link.springer.com/book/10.1007/978-3-030-01216-8', + 'https://link.springer.com/book/10.1007/978-3-030-01219-9', + 'https://link.springer.com/book/10.1007/978-3-030-01225-0', + 'https://link.springer.com/book/10.1007/978-3-030-01228-1', + 'https://link.springer.com/book/10.1007/978-3-030-01231-1', + 'https://link.springer.com/book/10.1007/978-3-030-01234-2', + 'https://link.springer.com/book/10.1007/978-3-030-01237-3', + 'https://link.springer.com/book/10.1007/978-3-030-01240-3', + 'https://link.springer.com/book/10.1007/978-3-030-01249-6', + 'https://link.springer.com/book/10.1007/978-3-030-01252-6', + 'https://link.springer.com/book/10.1007/978-3-030-01258-8', + 'https://link.springer.com/book/10.1007/978-3-030-01261-8', + 'https://link.springer.com/book/10.1007/978-3-030-01264-9', + 'https://link.springer.com/book/10.1007/978-3-030-01267-0', + 'https://link.springer.com/book/10.1007/978-3-030-01270-0' + ] + else: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-030-11009-3', + 'https://link.springer.com/book/10.1007/978-3-030-11012-3', + 'https://link.springer.com/book/10.1007/978-3-030-11015-4', + 'https://link.springer.com/book/10.1007/978-3-030-11018-5', + 'https://link.springer.com/book/10.1007/978-3-030-11021-5', + 'https://link.springer.com/book/10.1007/978-3-030-11024-6' + ] + elif 2016 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007%2F978-3-319-46448-0', + 'https://link.springer.com/book/10.1007%2F978-3-319-46475-6', + 'https://link.springer.com/book/10.1007%2F978-3-319-46487-9', + 'https://link.springer.com/book/10.1007%2F978-3-319-46493-0', + 'https://link.springer.com/book/10.1007%2F978-3-319-46454-1', + 'https://link.springer.com/book/10.1007%2F978-3-319-46466-4', + 'https://link.springer.com/book/10.1007%2F978-3-319-46478-7', + 'https://link.springer.com/book/10.1007%2F978-3-319-46484-8' + ] + else: + urls_list = [ + 'https://link.springer.com/book/10.1007%2F978-3-319-46604-0', + 'https://link.springer.com/book/10.1007%2F978-3-319-48881-3', + 'https://link.springer.com/book/10.1007%2F978-3-319-49409-8' + ] + elif 2014 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-319-10590-1', + 'https://link.springer.com/book/10.1007/978-3-319-10605-2', + 'https://link.springer.com/book/10.1007/978-3-319-10578-9', + 'https://link.springer.com/book/10.1007/978-3-319-10593-2', + 'https://link.springer.com/book/10.1007/978-3-319-10602-1', + 'https://link.springer.com/book/10.1007/978-3-319-10599-4', + 'https://link.springer.com/book/10.1007/978-3-319-10584-0' + ] + else: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-319-16178-5', + 'https://link.springer.com/book/10.1007/978-3-319-16181-5', + 'https://link.springer.com/book/10.1007/978-3-319-16199-0', + 'https://link.springer.com/book/10.1007/978-3-319-16220-1' + ] + elif 2012 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-642-33718-5', + 'https://link.springer.com/book/10.1007/978-3-642-33709-3', + 'https://link.springer.com/book/10.1007/978-3-642-33712-3', + 'https://link.springer.com/book/10.1007/978-3-642-33765-9', + 'https://link.springer.com/book/10.1007/978-3-642-33715-4', + 'https://link.springer.com/book/10.1007/978-3-642-33783-3', + 'https://link.springer.com/book/10.1007/978-3-642-33786-4' + ] + else: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-642-33863-2', + 'https://link.springer.com/book/10.1007/978-3-642-33868-7', + 'https://link.springer.com/book/10.1007/978-3-642-33885-4' + ] + elif 2010 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-642-15549-9', + 'https://link.springer.com/book/10.1007/978-3-642-15552-9', + 'https://link.springer.com/book/10.1007/978-3-642-15558-1', + 'https://link.springer.com/book/10.1007/978-3-642-15561-1', + 'https://link.springer.com/book/10.1007/978-3-642-15555-0', + 'https://link.springer.com/book/10.1007/978-3-642-15567-3' + ] + else: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-642-35749-7', + 'https://link.springer.com/book/10.1007/978-3-642-35740-4' + ] + elif 2008 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/978-3-540-88682-2', + 'https://link.springer.com/book/10.1007/978-3-540-88688-4', + 'https://link.springer.com/book/10.1007/978-3-540-88690-7', + 'https://link.springer.com/book/10.1007/978-3-540-88693-8' + ] + else: + urls_list = [] + elif 2006 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/11744023', + 'https://link.springer.com/book/10.1007/11744047', + 'https://link.springer.com/book/10.1007/11744078', + 'https://link.springer.com/book/10.1007/11744085' + ] + else: + urls_list = [ + 'https://link.springer.com/book/10.1007/11754336' + ] + elif 2004 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/b97865', + 'https://link.springer.com/book/10.1007/b97866', + 'https://link.springer.com/book/10.1007/b97871', + 'https://link.springer.com/book/10.1007/b97873' + ] + else: + urls_list = [ + + ] + elif 2002 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/3-540-47969-4', + 'https://link.springer.com/book/10.1007/3-540-47967-8', + 'https://link.springer.com/book/10.1007/3-540-47977-5', + 'https://link.springer.com/book/10.1007/3-540-47979-1' + ] + else: + urls_list = [ + + ] + elif 2000 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/3-540-45054-8', + 'https://link.springer.com/book/10.1007/3-540-45053-X' + ] + else: + urls_list = [ + + ] + elif 1998 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/BFb0055655', + 'https://link.springer.com/book/10.1007/BFb0054729' + ] + else: + urls_list = [ + + ] + elif 1996 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/BFb0015518', + 'https://link.springer.com/book/10.1007/3-540-61123-1' + ] + else: + urls_list = [ + + ] + elif 1994 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/3-540-57956-7', + 'https://link.springer.com/book/10.1007/BFb0028329' + ] + else: + urls_list = [ + + ] + elif 1992 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/3-540-55426-2' + ] + else: + urls_list = [ + + ] + elif 1990 == year: + if not is_workshops: + urls_list = [ + 'https://link.springer.com/book/10.1007/BFb0014843' + ] + else: + urls_list = [ + + ] + else: + raise ValueError(f'ECCV {year} is current not available!') + for url in urls_list: + __download_from_springer( + url, save_dir, year, is_workshops=is_workshops, + time_sleep_in_seconds=time_sleep_in_seconds, + downloader=downloader) + + +def __download_from_springer( + url, save_dir, year, is_workshops=False, time_sleep_in_seconds=5, + downloader='IDM'): + downloader = Downloader(downloader) + for i in range(3): + try: + papers_dict = get_paper_name_link_from_url(url) + break + except Exception as e: + print(str(e)) + # total_paper_number = len(papers_dict) + pbar = tqdm(papers_dict.keys()) + postfix = f'ECCV_{year}' + if is_workshops: + postfix = f'ECCV_WS_{year}' + + for name in pbar: + pbar.set_description(f'Downloading paper {name}') + if not os.path.exists(os.path.join(save_dir, f'{name}_{postfix}.pdf')): + downloader.download( + papers_dict[name], + os.path.join(save_dir, f'{name}_{postfix}.pdf'), + time_sleep_in_seconds) + + +if __name__ == '__main__': + year = 2022 + # total_paper_number = 1645 + total_paper_number = save_csv(year) + download_from_csv(year, + save_dir=fr'D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_{year}', + is_download_supplement=False, + time_step_in_seconds=5, + total_paper_number=total_paper_number, + is_workshops=False) + # move_main_and_supplement_2_one_directory( + # main_path=fr'D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_{year}\main_paper', + # supplement_path=fr'D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_{year}\supplement', + # supp_pdf_save_path=fr'D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_{year}\main_paper' + # ) + # for year in range(2018, 2017, -2): + # # download_from_springer( + # # save_dir=f'F:\\ECCV_{year}', + # # year=year, + # # is_workshops=False, time_sleep_in_seconds=30) + # download_from_springer( + # save_dir=f'F:\\ECCV_WS_{year}', + # year=year, + # is_workshops=True, time_sleep_in_seconds=30) + # pass diff --git a/te_u/paper_down_load/pdf_show.py b/te_u/paper_down_load/pdf_show.py new file mode 100644 index 0000000..57e74fb --- /dev/null +++ b/te_u/paper_down_load/pdf_show.py @@ -0,0 +1,9 @@ +import gradio as gr +from gradio_pdf import PDF + +with gr.Blocks() as demo: + pdf = PDF(label="Upload a PDF", interactive=True, height=800) + name = gr.Textbox() + pdf.upload(lambda f: f, pdf, name) + +demo.launch() diff --git a/te_u/paper_down_load/pdf_show2.py b/te_u/paper_down_load/pdf_show2.py new file mode 100644 index 0000000..fee2101 --- /dev/null +++ b/te_u/paper_down_load/pdf_show2.py @@ -0,0 +1,64 @@ +import os + +import gradio as gr +from gradio_pdf import PDF + +current_pdf_file = None + +with gr.Blocks() as demo: + with gr.Row(): + with gr.Column(scale=1): + with gr.Row(): + # gr.Label("会议名称") + conf_name = gr.Dropdown(choices=["ECCV2022", "ECCV2020", "CVPR2024"], value="ECCV2022", label="会议名称", show_label=True) + conf_button = gr.Button("查看会议论文", variant='primary') + dataframe = gr.Dataframe(headers=["论文名称"], col_count=(1, "fixed"), type='array', height=800) + with gr.Row(): + look_input = gr.Textbox(placeholder="关键词检索", label="关键词过滤") + filter_button = gr.Button("过滤") + # up_button = gr.Button("加载") + + with gr.Column(scale=2): + pdf = PDF(label="Upload a PDF", interactive=True, height=1000) + + + # name = gr.Textbox(show_label=False) + # pdf.upload(lambda f: f, pdf, name) + + def up_load(): + global current_pdf_file + n = r"D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_2022\main_paper\3d-siamese-transformer-network-for-single-object-tracking-on-point-clouds_ECCV_2022.pdf" + current_pdf_file = n + return n + + + def load_conf_list(conf_name): + if conf_name == "ECCV2022": + root_dir = r"D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_2022\main_paper" + return [[i] for i in os.listdir(root_dir)] + + + def look_dataframe(evt: gr.SelectData): + global current_pdf_file + if evt.value: + root_dir = r"D:\py\keyan_qingbao\te_u\paper_down_load\ECCV_2022\main_paper" + n = os.path.join(root_dir, evt.value) + if os.path.exists(n): + current_pdf_file = n + return PDF(value=current_pdf_file, label="Upload a PDF", interactive=True, height=1000) + + + def filter_by_word(words, paper_list): + word_list = words.strip().split() + paper_list_filter = [p[0] for p in paper_list] + for word in word_list: + paper_list_filter = [p for p in paper_list_filter if word in p] + return [[p] for p in paper_list_filter] + + + filter_button.click(filter_by_word, inputs=[look_input, dataframe], outputs=[dataframe]) + dataframe.select(look_dataframe, inputs=None, outputs=[pdf]) + conf_button.click(load_conf_list, inputs=[conf_name], outputs=[dataframe]) + # up_button.click(up_load, inputs=None, outputs=[pdf]) + +demo.launch() diff --git a/te_u/paper_down_load/urls/init_url_ECCV_2022.dat b/te_u/paper_down_load/urls/init_url_ECCV_2022.dat new file mode 100644 index 0000000000000000000000000000000000000000..a34b1f1e7bf455baeb4f32867b99256677b75e88 GIT binary patch literal 2299323 zcmd44OLH6PmagZ4Cuh$+$I-EScSjT?;!0Aks$=jjiIi9bOI4Jkqaz8VfD}O_3P6&g z!h@dq2YBGuY5cj0ZnF3&+oX{8Od!{V)vr zdcecWa2`~8!(ciK=RZ74<{zsY_L=Ohac~*_@a!rIuP5xas0rg`{ArNjVALb9xq+} z*I$2q;Ilz8AI8%^{MaBIhtptQ^vQWNAB8{eZoltT{)ftLGL0u;Fs^LJmy={3PAl)D z8NW-VGn>V|DA4cq{R{Uk?k_gtUsRcfqaU92f^j^K__sc*45#78AD#{8^U3Vxix-1% zejdm3*?byIp7-L*7o5rLMe}*aTI>8=s(q7JW4Ld?(t>6I*qR@!6+JxtI<5XoK<_=30Tae+NexESNTA*a(S*N zH`R9Kd{Dg}^11YYxS^JQh;E}G!b}(M-hH2+wDb<|4O`jocX?A?t6Wa?2q*|}&B&z*jdQrIy$BF%2vb!M~_ruT6C&S6J9}nWo@cS3R@=y7i3;l=S zJWjCqj=p@0>v0Sp!)Z9~g_z0q?UQeD?ch3?_PK1wzJ7}Qup=+yPO=xzwP%5roCZw6}roH+!I}@l6ecW zN*U&3|D};`%kc~M#a;K6y=1>(b~!Hi0#m;=%J1>Hi>FIZmLFbn*H)%6;q93o+P(4M zjg1#cB2Md`e_6h@?|$c~BHt8FtI=gJ2%lA!ely+TOFk~Usom$@ZGGzj{+p$Hd~x$G z|0A-0XZl+9WywF82U~Vh|L*%?otU}(o)#Wv%h;j#x4Bq zv*$st7f<_yP&}#S4bT4W&Z*k>*j4Lm_4?hn&Xf6^IGp;L#g(2_;&E?8KO@^KnpR`%jvG9w|Y%!gs+muTc11}8oL@%ui>-}aD< z@FwIB{7XFku5u8B)qS44<`q7~2lT1`KKtGN{JXzg`ev=sZWDjjYR@PAkAMD=|LAu6 zd#-dFTov~fP`>iY)mbvZ@96S+Gztki^I&>YxeU1DNj=k_=nVXUAH4Xvi(E~=(ev4a zJarIGEmgTH9Q2Z&ug;q4w;@op-P7XJz+J`Zp_k z#O*&g;_6l~3;UJh7~fXe9>qz&auN>66v=m$wSJ7JmG(~M-T9}mH|Jxsu(CUw<4^3} z)ErL2V8pRc?-|91VJIuYGQO)EMqh?;Qu)Q6oF$e0B>t{)6mWZoNu?W%zFY?5vZhko z_+wM~HYbM7c{;ILef{CPvNNvil3P>z@TGD(4aPH0=aPK!DDHGn3=!7ML00*=ED;+HV0{F;1M`BhIVTR~LvW^Jth@dOdkbP zSyJlYcfvV--XMxQZo|Aq{$-1}L zzg}Ar18I3Pkmg$JVFT&(Nl4@pw&Ut)7*H*7Sc>#NkI90<=~bkh;MP7|8h9J`b1sA! zGLvgKk=r^LabPEB(HM_aIlwf&ZFX2&YkM*guDl{9vhK}9+N}a-au|Wu-6SLDzl<*83D`jy6r$E#^JNRowH2|2O>eePTWc2J5FGtMLgjxCTKL%l z&(UB0c}4F3nm-t_MsN4wXmCu=d(U93m-zB=JPj^` zJK&b;l2_w?%$G)c?f?D1(hBxAxy;!TxW%U^taJ^;*d8Y1n}p*hZ+PQEh;qzBJ$xI5 zq!7_;n5ZtXgyPJVC~?VHvFR;u*j$yoq2>_)Z8Zw2k#0EpsCy-=P+jCWg12Bu*cJVG z^E4G;(Alhd5>BS!Y*CJM)MyvJpEH!h}@6^=haQy?P40&5+*`W<$nAu(wj{2_TJrKPm1>%y1M1N=+}L{&$sESD475T0~bR%>>qsCo`$17|54s^2D7oe|Tzby;0P8 z!M$)|=j4p<1WtrF3i6r0fgrD~b#@z$gmus?B|$yJn*hjTF%&xVX#z4%2aV)x~(%;txQrZ8QoRGRNKO+wcUx@iw$V_SKl+ zLvIualtatl)T3lHk6g3nnC=anCGfIdrAsq7B*(aeu>TT{e~Z$Lcb|xKql-BAL2ErW zWZgYgeEO&a#yx#=9gM4IVOV+dT&?#(NsDRw%3bR1mC0TF0jaf(R_)B~Dky}Kc$a(ZlTpqaz5osWuTGzM%`D~*{H8dcH$3Mt!>r|YVFh5n0{momn5>> z^)vNsa#`&zwJ?1><+WnxbVCwP=&G}7io#mX+`Y}2AKlfV${yS(H9iVbdVmV4ked6g z@)b4tRB5MGc6S&NA#qB_Wpmw4Z%NGN$|N!VP}llet0;-N3Sc&E2a}hT&R7H<%V#?F zs*trleVtfo@2!Qn`s%$n%Pw;9-ScGXn725Ml=?+L!X7Ag!l|&9w0=6^@wMyfsI*cL z8)Gi7`QTZ2j;WT$WEwSJ*$r1-702Tbl&#ll1qi(>pF496W{3nvMoTTELR#z-Sx7$^ z-L?T2814g=_!3Pi57dI#4MiAKA_FG#Cc3Pg#R=IA)^ZAc@ZbhW@7=ImaB4JsHT}5q zir7fq9}rxx)eEqg{V+ZWKT->kzzCLE3NdMo91Dc=siMM!2`7Fkf0Aj_WGcndmT7@^_QDbF@M|FQ_ZoOVaD|HpmrThbDHM_uB z98be495hC`xYK0pGLg8X@+66#gtC6h#ihjBHP)D z7=fwo-lP=BZa9!$mrv^y1ECwi=W4Thwq zlC!kZtk}k;FZ`A(uSyW|hu_v)?Tv?v>cyafW-QOCK;}1&m=vEUG<{u+aarsZ!+A^8 z^_+uIpQsA9i?bjVc^zF+EaPjig`MPPkW{wA8@d1%dr|_dY906R2%D)M%v37rTgthb zrZl!NiQ!(@PwknXB+El1>@Vs<5M#v4zmxiUPZx<8b+zP?dZ2a%7w zecuJ;jjzHHgWMo$-J3DU8W^0z^2{U*p?NrQN#dKtRmUmEVNltVBT>HMQW~qADUK)V zJ(_8p#StV4W@6^}q9djm=}tjw`WW6Aa5tS}ZYR$x@0f|!#7!COuG#cOu;$9E;%@xG zy+EG@4TA1P2-Br&!}(K6Bzt;Glg+u9E82mwt}1!3ItXBSE+e)J5MSKhqJ@{mC)>`EC-R23@IrMf1Bsm3l_GBsPiz!t8&Dwg68<*je57c>TJSZga9(+PM5 z&KqBG-0A5rH5KUuERr4j4(m=&BZf*Tm+~=5&4%bVJk2zYf2(W_gE?*H^J##-!zI0d zVZgXGCDnZuHpbK~7fD)mZ^lty8RPJW z%r+2uco-_#VswjWk9T+r+m6eiC3uCg4E2lH@t%npNzX8vuNWIug znod^x&uA+D`p*PAI>o5GQ0-dgVqpq%q#r0BBc4CsRMOA1eHFf~_No-V4S$eqqq$j- zv39qs*M~qR8Qx|(Nzhju-C!>a`_;X8Dq_om6d6`W#ex+C9*Z%+Y_b{r5I7nIi_2OV z79I4JwK#``ZHZaH=So_~y0<8V_6rR570{e%_(NeEt+fJag1u2pvt~TdTyZ>>Lcu~{ z=rrCls;p(I)h_k|p8tggqN7Wu){;+7I^+H{3{)T`i1!FtI|wcutH3;pW7EQon_OrH z1gu`vTHY#YAe-Jo)aJ^BD1VSE5vqVzu(Nx3+BwM#V+jB;5vqX96Q#&11EGAu$Lwh^ z82LD}oRj@6SLfuSkaSI4KNx3nqj+Pruw&c`9Fx(=YpCUSkVwB^OO6JGBBR}=p$cEq z53;x=SCk>Pzi0Z~9vZ4GaUNOz65-Ai|Thde- zzWT`R#;Ra04SyJHV`HtLaz5EQu4)44CB45{)EB(66(>+4@fDOks*|OECYVJ@@-~&! z0FkBYVW_@PP8BU@8YieBEZhbSVt`wv!6~8TP(EjmSjQV?Y|}ni?Ov!Uf*wqT`0^Ym zH)SH2?|NNQqO0v~CWI@mN^av1t8Hv<78ubjuL;ld;E0&3od9$xcr-79J>`rAPUIK* zHfhA!2@GN}T^%c&6uS^rG8Yq(#RPLmEmoSNieWE9xV7qto?~2+*WYgY%5k{zsu+qt zgtxg?FEW&1Qr(*}u>)&Gu7=^VrTXV3xvY%|nOosK5@u-s%-b_Eh(lrs&4N_Ua(3}r z(|%m(OvmwvsX`K8rl5bh(G+TzX;prhsY@%Zi6oM7855AED3VtX5X+$E$BsG>xW&(# zeVGJP{1D%R>bYRSS+h$yxLm_8@`|{GrayERI!!^-M>FP3eO}F;%Y8>03tJ=^=^zh5 z#ewu0)@IQ}nQdo$BiwE~x!>T4EP$l48WhsZ4N_0XpAcH$p6!guX!&v6=)sh#2cI$F z_BQ9iZ*%k<*EB3^GTw$X)>mdoqv;RKZPuFwCUbOf!VD@x97$!U-%_)t`U|81%>i0x=X2@t zbk4HjrzWv0uZWfS>)JFM^#ZI6*drBIbCt=m9(=|@%mmk56}!U#=^c~G<)Ar9FQW!z z7?}(LuacU|nUSI$e;Qknah1tZIm*@}*l6X-2Y!focYi&U=6bzAG3(uwp*V^yN7m5L z_fl+6wG~4b-jCu3hN-&A6d(!Bq&Eofp$%}D^vM4tffTdu;Op7M{7u?^X1DFiD{^=G zYl<{C)(YY|q_5dT_#?iW4%+p-N>Mjs^$-=AIexxyT|@2;T@uXhGOqhcU2JGqrpH>-n5RL}uLLf*Km6{}}uc6cFrM|Z- zugDGTufNdT*evL3KHENccd|=O@8~z`GJ=v>(=r2!cYDj59MW!QiuWxVo6mB$u+1k9 znOpo-T88kv;d9aNqI*HtYK)EG-uEGWO|{`=JgXitM^etk*Xn+FGgwIdsd+P5Q1{n7 zXl~XDXfp=t;4t2YQ#RmHx6OpsBGrKlxyHXtm)cKYJ*o1c?S0vc=Au;OOk;N}Ko+q% zxK#IzAVYyQ4Zz^EL~!>|G-PdE5BwKJ!?^mE@w9UKW1Oj{@glqOiue44!YQYHJk)DR**Z%(qwFW1*cky&A36; zyS9?^11P6UO7~4Ny3qbV)#$XSXbZ$cI97mwPTh9zIr^5#Ff7$x=GZ)`B-@o&#X9{ElxSK~GSy*n&C=Ng0Qm`7AqDSi_k&5cG9*7#%C=ax{XAcly)T6eHM*Dw20$CH4 zt{poi&rz(R`QX$Our^6>P#CKnHI^+M+1QH7Cc@idkXoxjq|af^aA|y(3u1Lo3d>s0 z`qW#d7D3D(riqSALHV;C&*nSZ*3Q(!iqjkf=FlQCXH~aN3G-hIG7Ju>y7M3IP!}+^_0y@2bDiiX-8{RIT@zk4sh6pg0=2H(5ER4k; z9NJo6FUWYFb%&i}P&dZq+1nh=VJR#N{kxRMDUqeI&|2O!p(Yd7mnc8_&af9_SiNHv z-etNhHIJ_xB@mo0gMQ^hat0pzM~I30vHl@9>~C?@v5Blnty|LLD9oy+rS z@}K?*lx`SbGL`TKyhRwy2f8bBY)|8I+EfHfZ$F5rvr4+7wjXgrEUcZ11w9#cd|_e! zVBOY65kpV;GRvh{9&yX~Mk!%=P%ee$QUgP(tjB=KfL-DsZHryXc_{agaX5+pu#aS5 z(*sU>OivZ)&?O_5y$K=_$PYElLKEI$dzM5A1TJYY8@>SE`o>dlTwPdwwAn%V#tIytqV-3}cwPpFWSmHu~V+7)r z?Pxj<+LzPN#O}1zRWp_)LsX;Yt0l~pSH*h#LBQ>`dO_#$Bn(oyx_9HyEeDtUayrs| z5R$YwZSQ;xIwb=lGZ~V*BqNL!kKzWS`ecV#V z7d%k%jXr#ki@d!cKsO&!grnvwYZ3n+-?SAZR8>g)lJLBJPKhROvOQ!_U>>un@rsvE zZ+h`{)caD>S{@_S{A~yX9uHVR-?RL8W?Oo;u9V6_<4R13-X!LWNYFZKT zNJe*3ePukr@1_aTEFCzM)|m!LYu|BiPi_vXkHdeYnU4a7`)D?B-=W9GigIOPw)o;R1kdS5a`sePI{Q{i z_337lBr;{q}bBWI?%_-GzMq~+++MLs|5%yrss*z#x4XS2jh! z(sD4>5zwwCeKHr}!I~YE46MALq);={ow5$27sJbt>R!1-zq#RSZi_3giXZU@12)=4 zZHJ}iWrZaL)4QVxyW^gQ0tPSdYDc^MZacInaS==Ql*S=rb7(r@U9UGXp*C1OCIo>( zO*GMH!+mgpPL=G)bsi+!Xu84}74Am$q%nYqXsLx^c5V$1N;qGn3F&;0I;l(;VtZiG zJ$hn{tMeu$h#nft{gBVANF`w{XOk^&!wy{8+2oTKpB{CHo9zN*`c6Dsh!s$A2z^Y~ zlY&$-tWKXaOLYldekJVZq4?d&B23-uRNzR9mL1suA1fTFYyH~u66c5djU=$CC zw-=J2eK#c*q5ww|oA>26xS(xp~wRnjGo>cp*$f;qgC@QSXs^}m`){g}J4-byCU z#2QPwDz=;@>!{_-<^-G_6|I>@sy_3GJ?kV_b9A_5Ap21~9qU~`ddcbW@Y;uf z;tYE#IrrZ3w1q`GrB?-9^{86Bu~yJmO&6I8TO}6JSD8AB;T;po?ZD0ARnA7vNS#S! zksBpB0kiq6b(D?-X7hKf72AD|K=ElnxPgK}=pltZBL%h`S)vMJk6Aig(#3508svoW zvMLeC9|pWxFX+d?;+e9{35-2O{H&)7fLzk}DLLDx9;;`4nmV-H;l);^@%ixt2GW zYM!QxpcjQweimG0Q6RWhFA@k&Ljdr5k5)P(EQXQT5fwQ8KVSk= z%vvNc|IU1+A+z%HQrf~-l$?KRds?AVU*pQF;@$m`Hf#0uf;BpOTZY+dkHtA@{vAsA zaz1c7t&+Q2Q3LpF#32d#pU z5b4D%w6+{^2Y^h&-pRf6;54X$VL190ndDDAwxK0W&DlCogG%2IUR+PJXAhGcr{}u7pl_3IS?2L=Vc7{TloYA!Mp;=7 z7NAPwo{p4g!6x#4$i|(SwwvUui83Lj;Y|$KuKe^xX1iW(c@zV#*O~>Kw8P*uJlO+( z2+oNsvX9g7GyBD-WFh0lkXoQEYV=tgT9I=bo!Rh$(?5!)UxJWKf{l;JgfK`%cNh4Z z@cK{Lbf#~iOWi*k%R-FvqKD;dq%D{Ra&NW#4#i9^m1q}t&sQrp8w5<0~uNF`V- zGVPZ<=$4-$cMiIh4!KYIAOHNLwlcCpknAcT>Auvm+$caTG5GQCl|e;UBvSmh(u2S` z`{5Y`+zW2p(qv^jXn&0HI<1?B7Xm8^(2Dwa!!p=(i#YeFdld#M*t9z9cIp`|HFxS~ z0c$)Nl<)VW-!#=Oxu}GmDDtu3&yX>IB$ZVhQFiZF~4SyYsdb^0gW0>bD1Bz_?6`6)&1J?;DqIwda;}7msszyfyt*{wJZMOOh zaSeY!K38@3RHbB#c2ciH1Pn-w2Lt)#T3SA)Ka#gnY3JVo%xSL+pK6o zN1rxfm6Dq`IoZ0Oxi!1;intGd?Uedvy#UWRK9S9^$>35+7YddsOY5H9FJzkjW65J0J=Kxwu1Aa)?LkLi zx+V>FxjdYiF89$=l6t2fOy;iXs#GH_S`W20(ejbS+6!n}eU=v9=cgY8h#b>db+zh2 zsOd-f5c+G^;2`#xu?NS+5zl6%Xp;q>EN8J2)}S(duw)!+`GLtXpVkV%Ol-I;ygU6=OQpmMk_OF+E+g3bRgtA#x2CmD2R&ZvJXg8~5wo zZiITe?7|T}6lGEb^h3y22H~NWy09QzfGF??46U;eF*g;$<@`sz=7-&hWLqs|cRo23 zQu7CoHXG{&LbYnN18#D9(ofidPd?HHF1N?0t0hfp$fa3}%m%b%IiKmw=g@}owXVvU zPs&UPJMtIN7ukowAtC`El(mM&BNEkbgJjnHL7B~F(e^qNl9C=^W?1cjNQRZ&zS4Y3 zLspqghUueB4u=C_nh|K1(?iG_nZ2ymwC30!j5sRj>z*d^lQ|;C=n_U@RnV}&IR#g_ z*OWV^g{o>4>eQ0=v?o1b?&(ZV`04LyfB0szRivGA6z>IZtw(xmFR6*7yql9d?a^E< zdgeqEHIu5b)fmIT5PnVQkvkysSD{!zT=MkWzEG=rdj+JDT(#g2!ECk~1=AhoosmqN zH+EoFgf>a2yBR;BS8XYhmg(5%@{apdep>oia?>_dh&=sRTE4=b#JPjrU16}W!WbhT zRHvh)9Xv+C_Z#_ge<&jD$AV^qn@uOBN-KM7KrM9|RJHw%kNyxcO%9#frB+A$!L|7w zuABN(L|7;=%Ew?zIRd3^F@AGHtavw?ePI*!1s~MfYl5!b(D^CRRQmfgK`iqkzAAT^ zbPqg1(Oh{|l6il)VzXVOx^qa{E}9(Unr`*=BB{6DL)y@UDb-K2d+zq%oe9N)VzVl1 zMd;>s=^06YlP>^YDrpOz7GQ(3)mC66*g1HAa3ZVq3LO06=$*QzX>@4}6vwOtJ?&F1 z?a=)_rh`bbW4ybB)7Y!1XAt2~%M32DHhjlXa51aAyPACp#!-3y;Em+cSQ(OwN2%QA zW}{#bcv0-zCOzAvY{uXyPpm)0K8OL1QM;26i(G|=dS1r7=CyPmKt3tx4w?ET6P&vx zYNx?`mfQre=H_hVyap>e!(1+qSAmdyLGEx60Ye2+U{1ayWeIbQmLFvS!S$=+Mm!4j zHaCkVfBIi8KC7vb7;+g?Gm9iWGkqC@GdhU;&WYuvsqa4}alki?)DYDT?OOYm#iC_UbxeL zD;@;1oAM$L&17^U6$!Sl>H`m5dlHg^gf^KA7nw?uNs2C-3akPwE8!@aVr=?R^TRqe zS72(#QVfrByDi27AI6Tmj03u+f}|axNK#0KeZ(FD?XOxknA^EPXD2^TW&tSkojo%- zM**Zjw;NO;239z8Sb~WM{3heFb!Zri_zwm=SV<}<$_B(vm6KRC^eJK~C?!of6KZ-I z1#hfz>YD|0Oi^tqSNlss0m!c<6Yz(dCCLCfQ;Rw!WoKFv8&23yl`|a1 z|Bxh^!Ti5Ghhm3BKL!co@>59C@s@04hOYmcVLTcH{c!PCN&Be#YFvW;u8Mv511VdL zqV%Jqb5BMyrZ!l=I%QNtOe;%T_Yt>JwY$x>F;~}6jP^(J4n&ODu4!b*ZUgaHo zCm8;je~D6A67mveeb77+be66M*^WU-N~*hAawkvjF>=rXzu|gUT@c_wx_oJ4;Ld=>0mt0+m zRP*)g#4(L^U+oaCyejU(A9C7eS<=I)tfj9fv_f2Ww7R5wW}akOBbKhn0!VRb?@5UI zQFsM5BAk9FBsY1G$Ga7~2=abJXH7(!{pW5q>)RM0trMZXXvo@65sUC{v3N#Jq=in} z0v?<6hs>{(G$U_pn&#%pL>+&iYkRXu1QrkD%%vOtmY9Mk0t=l5NdX zEyVXy$c^E3tv9j{K-PUwB5M~87j3x+CYQK(N2nS|w{a&xFM~+by%}9e-?%d!B#|Vt zN)8;hYQ7Q_uDl|_$fI2F`dYn!QUj<4UjsNk7#YSw#;SSc4%H6e0$y*`3PwxR z%K3Q*N3s_%hGyx=-K}~0fRYOW6K3RAt!lr@tqwv_gJ%mL@t6N|j&2Hpg#l+ANQeqg z#Ht=NU3uv{P{5gH6ZVp_=HY3@cZ1>QRl!p{Dgm>pslYtC$GcUNa@&d#<&qsN!?f1o zk*1EOO-^X%nG12PVr@&U>}tZOXJvFQ)JXl{Bc;$!rra8eH=qeuHfcF5MX;>ES~Ots zB>KT;BL1H8xudXTI?G1$KjxEs8*-3GA>Q@%wE}mKB37DKnuG6-Vsxznfe*}|l1IR; z&2+xmk4svpKdzoKQ-UaU z$O!(fWwVf7N;^Hn=%d4789CCYH8wH;o8@uNHOK(~yg>yP*IRv=)H!V3%0T}b;aEf` z6wXSzqI}f~U`9-=#lv@&OGeu@Up->3yeb)oKcsbYy`TwrJdH^L6h+Lq zN}5RH5pm=zI zr(W>~yw(tjc-S;F6o$7aqbW8DxsLbN;;SAK2NDQF49`>KDuI1syYM!j!0PK&i7Nh( z)LN}p1drYHb~^b5ZZ|zBBll?^0V^C|lvR_fbD&{N%0_D9MB)pN-+OzDX`d z!1jzYCu19LHwB)9cxnnR$UP1~dT9Ghl9nH5Ev^+Ao|=w=N~Hsl2ctcr}J! zjYdhPexEOzG&I2iKyqA)U7S-|LG4o}E6sKM0=;e$aW?x}70Hcyxt#5+ulWL3x$=s5 z9-nf<^;&@z_4biXM2$W!XDpP@i#kU}h1t0Db6u=g2Mh6SgM2?193F&hm}$ z4*?;t9|tlM$9I#RFl8w|8yNm4>;=lD=Ir7`+LM~aENMNDSxNp?o@c$bD)_5Usp7_F zLEK_e8qie$OqQxtT|lCv3=_b1_gP`>#_JtIGjg%&cWP*su!+<}d5j5YrvBLMK?AsQ zGfuEYJ3~`JV{AkN8b|=~n?SJRnVLbS?y`qfu}cAuwkXM2JL(4T%T?J;Z*s*zd;>#Lxr-diVm})$$pV-y za!KDLAu^(Nxj@BnHXZ@nmBaS4``VwdM!i-D?FZ9AIInijl~JTjkp%*q`|~^H1tppI z?>HF;7eKTO_8Qt}mdc%SFeU+ekx(f_oEEcE6r4P}B$^dLzKbfEjZG?Hr=70S)FY5J10WuA9lQydGWxQpE zL@p}5MMrCCJjL3l>7adQ9$&KZ(tSRU44g%(+ni+Up*_}00O~y27N&93aSL-%28ITD z5G0d1&NH+eYVg_4r|E6jk^2L1!yd&De zqDZHPI(YYrVmj?g<=i7Zq*(Q}xkDR3Jd~7~&3;T&N!hsp-=KMyB|*s-&QtDlJG|z@ zrDheIYc)@Gxm|fh;;FwjL%ms)XK6?%IA@~Jswf8`Yqv9`Q+MTa^kN5Z!nU^Q@66S zWZYlRq24Om8T5)J=95WCR%|gNb3c*;QR^U5#kB$Q0KYpq8xvcHXs_o-vsC_06Ov(7T!bQXgaWuw6QBck}%-QXZtTKVGma(eXFUjCBZ;68P5TSZ|&nkLm#XKLz?oGXn>NOwO9>ygL;<~359L+VZyehY~KNPLWqVk7HB%OY+7x$9c z`^HOv5RHg`j1(>=+^Nfu$=JumFMIIJ#+e& zU0@2Ax71hRG*ZeHsEet|>Pp%}(^vLTZ>~)C;1BUbGqqsD!cI84QtHTl@>m!GGRmmw zcW=h=grNy7qIEB36cd-18V%`p8gSiV@>!Tex})RQ#85leuXWA-RFhK-*IQzANgMIB zg@aJ^#6%P~DZB`Xq1A*_x#^Vp(Y;F{lC%;p|x~{yATOOkY8}U)7~D{En6d8>fXEchf-fX4;h5&J1OyoBq=4NWlIAiL zPB)op93?AJmH(?w;`Q`$>@@`wD+z5iYL6&szbcdE%FcqGO!;QvSo|TI3?F7x_H`-Hl>3gfGsuWvN{2;Mc%m^dDdCB000Ny5AFDAC=? z3pq(UvkB&Xp@fi5%4mA@I_2~k?YG-E@A(gt3z`xK4DAmBo8Qk;M8d%<$YLL(epJ?j zw|!v|jrOWw5gwIzTkRq&B7?5--rG8s0oBF>K70>GP!*LkBOU)?S47(%KM9r)MsbMC zfSZhpZ%hWjP`@ExZ*2U9xftH1@}ThGqn)vIS{+zM4r!Zb(`O*8CDnWJ2OGW^W?Xqy zJdHollwA=8;U=>v2e5YrY&st}Bg!UZY@`8fl{q_Dq-iT0x{6BJKgtXsD@yLcbv;B@ zH;#ntLnfj8AV6P2##XXd1{42}vx2Y788@F=`F)X&_(MP2Yemw$sl#Bk*GLTV4!8Ym zW&)aK7YnEY#*IatrQS?F9T9}SW#m&$&X5XF_D)qOCY;WrSkGxzZd zon0pK(DRi<5d~|yzF4`7C|pr5Vvc5B*+NO?DvI+O&j=}MWgv=F)z=sleY?wY`E0Y{ z3%lUTD^hvZ{o$eQ^+rME$#$6OH3Gb*QM1fu!I+7&!gEa{QvO*A`$y%Pt3VS?1WD2Ph zI(4y&$}i6=ThHyh&tk;Oq)?ZflRYT=q>(xZVMeo^l(#glbda2f$l$@YI(W|8Nm;*(GKFY*h2sOEaTDF1Z%#h%3k_|;l!?-qv~VRz&COrgMiW|k@t-Nul7Wh)fB zs+uKr1j;3l&~-dszLuRrYCeVq|5ZsQS6-DY!5=WWzF99C)VFEH_nqBzl0u2XC(~g^ z(=JsH>&(lx#HR7zV*jh*ZmvvDkv&SeCGmcg1`D~oKv zNER#_BFKxFFJ~|yVUTU0G~=KQS6o2mbS-xC8|CtWR^1ovf-A3zN%&K{-CQqFKO_$@ zEp56GG$U~4Ha=hJB(}5GAr_WD_OGxRNp_4X51Ppt>*y9G)_72I_awWbuZbc0hnT?~3s^vw}fzwe1vXu@?%LBbp zCrABKw~NZ5Up-bPmtBzEM9=mSRmfLmCR}+{ypBJ3wB9b7z)huYF&k54XC0N0iX|;v zsgk)m=rk0&4@`8)z^AkAlx?xoyYOz75wSl71Xa5JkU;X7~iwynsR|mDPrZH*un(l^Ok#{F;`h ztsOX#C*RtURiZvsaBIy1o^J=O9;~3+6-&b1(~QNWh=vJhFeM!K4Y=GIk$(a?eNJ-CxCXrd2&o8Zx| z`Rc84fU(3oec9!vEfX!6QAxnF{nz8)cvPuv)Qhl)4g(`V z4^!#~EiM-RbVAIB#V;yv_cD&96);G}0b%Q0a*D)&DUcukkF$mJ~cf7r3)+< zo5n`O+Xj<#Qz^G8#QmeT)3p?!b1@qRHz1)ysWNTR1!0#q3hko{)<5yeaWt;d8gn%- zVi98P^`|$z3j)wCRbh+Z{ZIKZD4s1 z&zam>$#U9qYbUZq*g(sh8glESoT6OLp42wg9FyV)L_M`b1c5M|;3D6e zdVNeB5Vh^CqA6)DZC^EQyS*|stv|?=@#q5H#-e$)`x#Nw%Mg0zVKf-dIXC{hRHdXq ziFNCD?k<6WSf;&%46*>TgJ4Y5!zU${&*D9|n5yzNxulhnvnnAXuEJGA9@tGb#Y6RN zQpO?E(QMEpxncT^$5KKR%b3Aez3N<+mV8ScwIp)VUiY&{sL2w^r`EB|a$=7Px{ZyZ zU707ZwmRD{E2r@_DEcfLV@YFL$Rb{1o?zK6(&2tALXb;rvUJ$i-SyZ;_4aMffDRI% zeGn5AQ)CWQ!`KhVeS{~H|6z;F0#ZThpZ#_+rPMhwQD<*j>5K&#Du?xD!{7{wxh zrTlrP;MVO8Q3S+^=v89d=scO03OyXGCp9p;@`{*@M`@9bMnSg*`^ z-E$8MFLIxakTV^`A}p;HoDUI^I!nono^{3$lK3xK(?tR-U`x?HrYr>zWfmtM2Uqd9 z)OIw^vf+oqYFAznEAiI>YHn^85a5=Lf>f1y$>}OD3NUqmIy%OcSgs~HLuiLPWlRc5(7SAg&*d9#vG9tYdq4v$YD)R&QCssBJ-G6UI1+!1`&J>-%_067f`I}9v)U3WmJn?hS<>Fn+(xEN zD~d@>R|x?31UZ?9qsUpH`c?dT>J{6nx&+kXP*P{2Y(#m5+>DY}eaMYniKp_LE+KtO zs$jQW*fHgPNj||}apPP7y`StNWoNQlYv{cFF}b!JRVDKb{@Ok5wW7g_ZrEcN5RlbN zD9)zX!TI%Kg2#=(cxca6J5128bW@NLO`Y~MI=5kPJ?Txg!N=%pxtJl)RLvd*2*xNl zI&sigHl;%q^B`ZR7-hCx6sa{H0g`^3BlFjDX(K=MaEII@57d&CHWHkRG z;riT-eUgkXXr~!eiKq&uAs(ulXsRuPF?EOfBuMm>E)Z+$k6;FUoBP?L5OBNMEGX1& zXWj$@v8=faJ+Ss%Qw&!W$^4${u#c1lr4Vzqph zz3cQiGxN=;Dc!&gPTQwCSnHpbTfjmb-S9ISZ&zNCJgw#rKWw*(@WTZHl(M_rU7QB% zR5jst7@?rP|245po^F&)F&awbs2WHEULimHoLhX@F9{CTJq;?g&>Vj1L8UB-tNB9* z+pVG%v&#sM^A7NHAUNDQ1|@S40Vl(I%00^G((jqFSkK9BhY>VTA@Orw#&A z_(VVEtdt%#Cs=<(9rul#ygxK_7NGui9(JnP)%fUoEAt7T-yWqoj7ADS!Q~B` zKwr~C(nbLc7HjEFBj(((9$q_nerq4po989FNcBxWEO30>=2Nr47oNx;EV#a2B>&0g z@KNltmAT|?B!~e&fc$~o-#w9N`?DSa!wF+7S<;~v=;wEM%i8NQRJ5nMnO$-#M+4cAet|26jcDAyE7*9sJy!zOP=?V z`9qw1s=Ro2+nJ3a>Jrv97-*0^hG%w%&Za`q@kw^XHIe?}?VdCB0L5{oBVf!WQB!^u zW?1s%9;1i#RoO$MwJJr5M@is%vt5uq=&X19gHZEjHwph{#|Qu zld7RaRA?s!FEAZZ0N*0@!)f3INqntHR%dlwTn8uRb^X{a6igc7tXh6EFbPYs03&t0; z5Dc3>K-R+42=tCIj`X=BV={h-KFdf*4TuMC@g_rUv@3h;8-6sDNDMYs1)8k;gC^J8 z8wJ#p+gjE(L5$}vDR9Z(t-Vxs=st5VpWnD#!4?5vn5V%kudswp5^^aE=^IQ1FW}5Z ztE=3jlK`{eg8gYg&Ttv@$R&6zV?vy79s`7KB{a9ng&9_o|1mk>tHKOdUXgUN?hh?( z7NnEM$M~ny_yQfh<7qO+4A^sP8WwkSTaV%~i}psr)L;@Pe`3SZZXJnz&n`4sm~Rhc zB_`7)Mm;UetPf_Z3eqv=(3&7`2r^EE$#uxmfkE<7QLD{Y8C?bM1~*n1i@7ttCu6QK zOIFgz9-|7p!1109=kv+z<%<^-`4`V;;5{r_nm+HvmoK;nCPd*EEHa3vLiAtPcFBy%{w&+7!OAo(i(-WqYVnd>YsMr`s zsF|oon#()sLVu2|7tLuL2g#_?Wkyvn1qp+-J5ePNF}^F=!eka!9VE2swaZJ zfxc0)qbR0~DR(pXA@_3-kG_;V*v2Db&Nn((e{g24h_ZfM?VcVn*|5wZvW)>kF+PY zxdI7mDJ=pt61{O%xbmt*0e?_syfJh*YQ~T{4x)mTU)Tc~k7)sXLi) zFcVN8J5{{d;>NXnGPX%^+L~#)V2@OLZim;b5Gt8WHocMCx$>$Qgg;oaT`S-&36+H5 zhQ3j+H=^r6Pn)r+w9hTSaYqXJkxt~NNVTS8#H3;q*b8Cb@tQkK@KIgR=_nhRT(^<^HO#0&{bGDr+847BA$(jE=g^1QDYd-R$)(YTunNIcX&K}wUv38zvm`jBb#Z&Y#9wCWi zJkNA^OT!s~w%P1OA@Ak<)PBG+)_K6$1M-p??-0SLa6h^B-2eAKVj6_FU9H;CzE-9J z09m7pc*KMNVEhz95u4}{c4Ha<3xYKA%>!TCk9K(red4FqE?;;ie^6|#T`v%)*gkx< zhb6Hjcj^efboC8CajR6^W1NbR7dDEKY{+mh7bK4W;s98sH$c-r zae}p}3jP<`{lqjr!m!wmR=KHnL<}q&#XN~?;dq{0Tx;QY{2{#9O@T_FlPg6GOLlWS zN(NQTqDtrZW-(u?>ir~y7;-nn$pJmQtLcu-+?vY5dX~{2kYMFGZo2K;*a{ehr|OI= zuS&+_4>Yc?HGU`K5&a9fq=k=RYmU51%8AX|0j?M1qSn0y97`dq(Cw&^2=pWCz{wbw zh$rMb=^KU#sbEVdmXWzlyUyKjFxA%fwB@h0#+6saf;^(e%_8n%9t1Rdc_?p5xzies z3M+4jVQLiiW)`!|7a?dmp69X~cq++-C<&;DCeZ0G0}`U(H?d%p2*=}}xfEK*R(y%7 zqSw12dGhpz7SvV{($6l#zK4MC%k!)ZwW;9`Tt>^b07!g?A&g?W&aT}&)516Q%~GDD z+a}Qi`1VNWP+b8A%N1x~9{IHa-|%J}Uynj~QI-Zd=aV)kZ#ReKEJ-E1g*$_Legc=I z#mpjy#C0fB1p~K(gD9vr25hDGos)DX>BHaqL z9c^@84QBq@3NYIv3haIr`Vzx3>2`aOmqC$HIN+YQh>HvnVy~B>gC^rsSjU{06j_Bi|2aCUlbVnrSo{!m>At zuhU6v!((!Hv`Zn*Xbqet)dPtm$r`9()HF!+(6U?J|G*TEDjon|QV*ETroQG9RBm#$ zu_cBsier_OMD3|^2DHBRh#C5?Lg={iiom@dg@ak_UQiQv!_mj;vb(t5CVgkZ6BAl; zMElb?nK)U;$7HJDCRhq97nDojXJ)s)Mn$v527_H7%9dY(StQ_^;s-|Lsi?ryf@;;c;OX{FfO@+>c$S-6(xUqQ1YSrHHFf0|#IQ7TK5tpY} z$_@DI=%77c5L4bqOz8$AozN0f&SkCQErh|ZmNRjvjF;(JgE;k*#3;6J79TztfUKQJ))QVZH~+%v%OV>GUyZ%sLngl6g+de zn|l%txF3zxX2ph~py1+A8}m(EN(4ItN=%gqLNWI#7j#rA9aM*i51aXBs3bOA=gyGV zCFAOzFq27U${=?IRt(7_WbsnGC2qOLsC9isrD0cI5f|>SiPCHp$vM+++hz&Q0uHC zd(Mxe+hyrhpHm*F54IK0=kaCr=gFf^6b;nRi`)D#8?6=sVam0VwNVm>CDUnmTwZFDObF*G>n=f<~ zX$NR=0-H-GL(8aL4xKyu_Xuu{!U&SwHh~gkhnqO0q9G{bxCf$K;%lvpcInw`=uSl1 z_rZt~EGpF>;HX-D*t~Y-6-l~0%H=jUi}VgaL{mp3OnK!(*<~d z<1m(Fh7Bqngh8KSfW-DK@1OzL3-igmn?_61DzQ^g`BQmC_LL=Cq*ny(l!kB`Xtna2 zYwdH_;vUFV45xeyZ|<9%#-1#^mkj9Z-mGhV71lBsisz3j-l{hXTF$_Bv&nz{$A4nZ z?fz2KIm0T|6Y^@Id$UQUdoxCh36Ta%>Ky=sC?M0l0Dy#^dsN?@8_h1ACR5~dpwmUB zobM8JpX)@8dtNS~4^4+lXqr&!S})N68M60MUM^n%!Tgc%1U!XI!| z!@v9s`!%U9)7ryYT8NQ|V}W)|sbSnU?H6CZiYvQF`1Cg94{L0-i)a)(VK_lbqaRWE z+?J)n7t(U58~&Cc9|MRW0VxhQXE8{6Au@w1m%Iwl0%ARsJc6*SK`!V)@S?yh=cL5y zNJNuoN#EqDH9c&YZWvDE&K6@Xtp-j3ZhavuatJSD`e41CFqLZX93^WP z0k$mU&otoXs2v$o&!Cs~A5F@>4#MkLo7$%rjO9b8dZk&Ul8{Nm(>@O@VI}r?sKD_k zmfUWy7j*P-7R!AB7ld#oi^xRYLUYkx-wCU?s#c0t=~lj6P&yz}Jt77PG*c*|z*)2# z^Ye=O92|`->!psSv+yptv37qp%vdt!tUrRR^;NZTV|`UHl)68{CZRgdPQw=obDruwXVS{lf-9XF^FArdfleP` zASPOG=b?@7kO_3(o3vk8(;xUCl3fU?2bd$lD}{DV7tO|6ZD0Hj0jr!pY1Vyp5V`V- z)IfEASS{K?50^k08w#hN=CGri>iZzo5sC&wTWm;>`I((jfH8t+QdKGjrWIaPYAdN( zo;k?57mPxDa#vl8*DOY=aj3|878G;_@X=mYBZ=yn! z1)Qn-!*%OL#JjCni3*2CKEvz@qmUuhOII5xswd$r8@)E{4#obkmrSXA=Tt}5o=zd( zoWVe|Be5Y5v4oiZqKVm=gMl$~4Weh9$&wb)_LZVE+p91+Yt@lIT$hcg54)75vY2v~ znIPx^4w0)KN6<4(W>cpN|4z~jDA;`M zKZQ9M;eiyLeoa!^dr8Y#_m-}#r|HV1|MAa1vd0C!nY2o9 z$Q2rBunq%5y#M}B{0pKht@@0vD!mb70zW*XdwxMl^KD4Hb$>wbMv;zPhV8RUERb}| zc07uq&F7j!)J{fO*!>K$3a>s*ewRO-b0;vXC#tE9(?W1TTnxKlkcJoajF&K+&e)8@ zb)_X}hvlIBpz6vSsMb0xMKn#b`6t4 z1sh>f13AEbRH>S@qe{DI#CLgk_&S|h6gln?^k2a38ETqXkaZbf2Our<2N-lj0D>L7 z5aufHzJL{j-3jQ0s{j&=N$p3;EEFWnAdmDYwDE8{MopTxv_$htrguH!Y3F>1R3}5@ z`$HO<<;<*A^VKil$}3V$`4pJ0*9!pPy-xSE`u-<4Ft<{OOAFGB!JxX_EUcq_nKL=> zRAOWr4D6XL-n$>bNQF%_NCX-AD3bsZXrrP^iqXoBdCy8K{g?38vGtkGW05O;j?kr& z9EoWTiGF1z7#s@hwxmt{bv7o-gZLD8uCEuhf`&IU?XRQyaeL8r`)*6BQNa6*Xs<%i3H(d`a~WLnI{_bxDn7o9shDInT7`8HmRbYeUTb+PmALYX z)Y=Vy*l(>-GzfUOyJK3JuWd~K5Xm8z((bY$k+z_D2LW|E;rWO%NdX7z(VpHN`a-#+ zMw;^0w$?0}`3!Oz8x@oPz+1>1&@DTRZ%W?!o@ARbw3U!;HlO2T zzwe&B#AKLu{3V`_fZ~E{nA?yk3sIOY*xW-nE*HkFL#im}T7>1=KpV)58G_}$snsYM z0kb2yC<*C874$@xS`AfZ^DF)!-FmH2U>FNlB^nZpZ*;|Gnt;QHPV_yf>~s%LD1EeA zYe`*1apqO`I14%Gy7Y5)(V6xMEPdP4djZ2qvWt@-O+s);Ob-R)a5VZP|Bo6Es$#}BUp?p;c-I+?2bZ`l* zaF{1ddwN#V+TNo@1+0B%R>KMk^AiC4&f_>h@aX=r2m=Um=;S{%HgUT+gJd%dS+X#hB zwmB%?m^gRod9d}Guey~huSiJoD5Ts#Nbq6e!1igKC1%S#-$fH%z?W%li4zf)DY$9@UJ+!liZZzbN94%^*p*l0@cs2i8nvSS zYIhQ$6|!{GOS*U*_oSW(;4q=*ep_2u&8hED_EE4W-j*+9%W$Mq0G0)R!Uk&tKkKm!a7mr8PW3}gbIKK|Lxig?j%v7g*uHmgQrUDPhZ9@gL)Fdauy5gN>t1t68 z#tzFm;w+R%&`l}?JJ?$$d$VKU<}IA{O&~1J)synFU3o>~jlV`o1J=^R3Fl7zV5uL^ zkOHuiyhpAFZ$--is#nw#$bKgs^SlnGUKMwZYB0C z5q4xxMnqJHO_bG8u({*-90pMpx(8EXC0)>lA50X3+E@Wh)Lf9imI)i;3$oWm%LHkG z$uI-TQ)^kyWwnqh%XCE7xr2+yFjULf5GW>qw6m;|9UgjYqS4Ma&ZPJ5mQ7JE^N=x6 zZC+N|OO=eTa%Qs=xXYcE6A@-1)S4d*#jd;}j>unMrLoyAV4sQ4eB7n_T}ok&hcTky zq3sL;W1Z5^|Jl3~pesR#t*M%$V2R;^GY`yW{wE){I&1|(#xYwAFXGT;!kj1QJHA9* zvXb=>3B9L`hAXd#S@`R)G}kr?8u_o-RVZQ@xCh%DXyteMpOV?!cw&M>7NZf_p1bY0 zZ*#unE-Z4U&2`2vWc}Q>W{(|9>>D-PxQ&~BPk{iDsZ5gi(2^HKQh$M4Ul%+F|}TwB%O)gm0-zI4PW3MZFhF?B>2%y+Vg^i#!W*Z=|Gq{4qImA*9;P~_I1b9G zkaO&wbSUh~&Y+$?k@1J#G8tb`>+HQcJbLNi*jyZvSkfUY2tt}c57-SWTxQ8724-t= z8OaPQZ%A)0U?RgQvwaAs!8xpCOzt}Zp`|R=7~Y0 zuUvGhH~kQkF_q>D5R+Zb>kqbV7Rr#j;RX|6TO=#Ae;ofb-T02+^ejdnc0yLXR5dfr zmyy(z+2mZ1urqTksK^?827lg(*raRQ)daH12fU3uO|~I-dCU9V4e(+$pu&$jyLuCL}VXyczy7lVk;w6zQ&2 z$4JqRrz9qgNw*~&B%@Itfx{-~6ge>VO^A!4ldbW*Q0sEhPeCA*$%%-m@qSd+>#X~# zwCn3DKy!0e;t!*3fnXIhi?+A-(n3nY)zc7PFpH#UZEk183% zHoe8L&6SB^{-E7fyUJi0x%qsnL!t8f-|l2I%t?#E(i7J=3}S0OZ;LrehrqR)^@LGki(us( zO2b?I#FbaXZaj)pw>CBlcCKti1B*jB2CZuu25b?R>`V&Yp{j&te#C!OTfI$H&MhS6 zGnak9ngsg*t0PspVY+W{5yuzOzM;)*fX+Oi;?mm(*TD_?eS|m+14 z=t@a4C0es6m)ztv0CV{bwG@V63T@5=3?N3L3yU}iQA1xg7ZEexT2IPM5DZyXR?-!< zePM%*_Ns(1pAy%#W`RtSH5)Q5H=5MA9f?+G_2e7^72f1;ck11vzXOcP7*}ab%bAFr z%*TX55PFhSS3^>xqvxPCbIpfIKj#~8bsOt`ra5p%>uErGazM|gP<5?nKv2?+=`tUq zsz76~Hnq{a(U+c+t^frO=se|vPB#F3A>q(^4ah0zT+I`-A*D2HFfU*Vxou2Ri=f&> z+o8(zrW@XB9Im`7LBk)U+pcXC1dX(dL`K^StCYvqC$MZbf8Yli`_AwY_(&snu5_4# zDaw~BY0%)4M&20BJW(007(YcVtK7-MxA4SU;>xRX^8OIocD+@EsC}NqTG*6Q4zff6 z?h)c?lr#&FMEPf`4cmq9Ryo)?vC5Vpk&6W#cq>qz-s3q`nBWgeCjia5FBU9UUX|nb z2dHxV1*DX&8ws8dQ3DwaD*@x-5{ZBfpRKxZrD`UX$lPYYZqNCb?CfpWubZ6D%ym%O zR#WZ)NnuJxD4qt>1Yq?PmA$St1>FYh1A+-G{kZyC2tIF^d|H(RBE|Emb%QG3s&-*X zuhR6DQP8ko6|?e(z_uGjnAvH3sdQm7ywOV0(`0&{jBMOjqT}EgY-|hEV#Y|iy@--z zlu}ewkKFetZzFuqFi6NH&tk&L*>k>a81{x#$|WtJ?F%(+w$qH_>2peds3{guP+8mx z;B6t0xU?G5dc^5=>I@oPh$!YF39Ghdc!SD`FpE3*hGm>t-N_THNK-U}?^t@uGXpeA zBNpI*zEY5l6_6^jc1_(Mj=ElJ6vUuKigC=YYN!Lj%&8QCdtC*+n`P?bG3x1JtK=t4 zb9VC~09QW^X5%oF;Y9J3gy_N8#VoMGqB9Btg>+K@dlfz&3!7M8O`5*r$ zm&_7&eu*T3r;U(WEUF6Eb#C@x?lU}AIiBD2hAx0f!PCt0MDraXs=ql~>k!dy&^_q} zLwbANfDMxbHI|9eG}iM09Uu_aX>c9T=IJ8W$T){S0Gs=Kn5H*#god@fRMj+npKnU& zW>>2xzr~eTB+2n8M7^Al6~WTiq#D?&5r z<@b|zE`ZGDs-ESYz~J2$73g#ic%*WErsYX-Z=I3m71*U>Va%sw^hQC#bC4VnD3C!& zM)Rn88TWxknmgd0@&{`EBHn6{`+L}x;H6XT2oNw8%qWpQQHMJ&OQx=|C$D6TsC_r2_>l|h(>6^n zejTxmANU2O>J;%8zjIlh!KhW@#5l=nz7=MJ8*}_%8-G}WS<*&52?1iGs~`mEY}6kY zxZW;+`F7rStMB6xV`=bTUV`_>#2cqlJ&nm`W|^YCTM!jpBIf}G*vvLVhCuhlA(!pK zO~?g9(ehZd;q35Cxtz1jA;tTCf;-z1b6A#H(m=e8*|#_xE%R|9^NH5_I0Nwq1fv!B zFmq}lH!aul?s6Pe)8oFoG!CJAhZjkR_z+7_XycQPK28dhw+k)g(ZQzzFLGC+L%{== ztP4Mj>K^O(R_ z7|xl=5T@e!*)k7(m$Io$v37S9eFWM3a*>E|W#e*0&2o;2%?O^7Yp%R1j>sc0y;i`j zQvM8vIGv#4VcApV-qKULlTK|RQ7A0HxuS zwicBka|oJp0)@B<7{O(}76^g-s9&+v*zfvU0h+T7UGn^W^#Bga zT~Fa=0n%+=Gzq5Bc{1nLPu;`48$Z3%e+h1w`_K##S%X$&+x>s&mVAk`FLINmolY@~ z;nyt6WXb3^-LkU2)^7R&0=e?4-1Gi0K^x~Cfztd1Brv4bfSP7ol!f_xc@``l78OxMNm{ovA0s8 z?>GkAmPqQex4#wZ@~(Hi3#6qx0ia`-GBIMCUD>1{8^IY#>08xNd*S#0F=qbl{QDno z10MA@qAVwUCXf^!5X03D<{i97m~;5G+3?l9YVrQbO;C84jOLkw`sXC1@or__ZD=>TdIq5+k0sXz$^kX~K3K#ZZ+2!^8x zGfjBRD00DT^aEN*svi|(r`^|a5KD&M&E3+LLj_im0QDpg*SXjh)#?B?Tk_ zMB#625M=C)_rl>c65p~5;t~$A%+8$Zh{}hrW+4b4ZYiR@z&nJ+Z>3L|D_^4RfDYq3 zvfI&Fu01X-WDlQ^)c(!F=lt!j3aP|zR%*?rom`Nd$fMGB1%}S!h|Jc;#z%4%)Mz5W zm+m4TfH7Xw0{QNGLZk;^Ka zv0g-{ym7d3m@;7cR%)YPVt1M_!GOLbvjjGZ>nbOZ;;DVcW#!V4AYOtN9_HGUyXsBD z+fFWkdDY9S9yPoxYgv(W|L|nvkYWB4a6AXu9G-n4qDtBM+~k@Q@J54_xDE-i?witG z%0Tu5W-0+-@J&-i#J5p!IT%STWIEKIsa=*R&uPX(2}k~jb#`ngizd*Gski&lh(u}q z6+tz$Pv?n>TxrEmx@{*H#8~`kepjoj8B|dsm!_&00%r(R-Gc&^Z5t?KtA-e8oJ8fV zZ&IYOjzMfm|By1AQoKv)nq6Juq-$A0A_IkCJ0j-fx>JVi|Lb7?k}sNnCTP6qtU_>73w_dJmgKhSaEU_~FZ!YdVx4u7Ym=CP4jV7yS6F)ESD4d=lQy z6{_<~JwdGm{zls9*6EdAeMgCKNHoKa>{`qH~h|5>gD?N4LkO`D^afUtzwJ6HdC!t$pQg&nO4iv@LTknxF!ad%E`D&l8A8S zcK2y-@D`2}PdAEqP^MhbN|-^A%pQMXBw!=x{!AnSe?1j0BYit3>~2Bi?I0{45S<7t zDtkByu-H-)y^tL9-U;Rz#hA5d1Ry3Jat%duzttxsEiZGw{+dDcN+pAe(k>sq`}iJq zW%~}`clK{ zv*}DrUYu$g^$hf?+6tt!fVdi< zZ!;mK{MMDhv5#us`HhEeDPQ^HOY$50OC-Oqx)*=@`ya)oGjz?PB(LEVU6SD_2AhD{%x^KO1wrgoxFpyv7W`%ql7CrxlWxgZYArzUu- zmch6aF54c%!5!HkrK8Yjn>W5gel5vwWLt|iIC5b+eZYE2k-WYH4Chg=lG-FY<@p zGV7ILO18&nJwRR;!(sf%Xy5U2qo8L3(XyhrFNA$PFcDhm^a~Z9U@lxURyTy$09A$x zM_LbT?F%NBKpil~Lc<&z(Ta-p7b*(ck?)KY@s=*=WYFs&5@C6k_!8=pL-bTFjz5BX z6AZf%s1W;r^~b4Wi6+J1>N0XEI6#iT8{=X?=%MQcRdAY5Cyy~7PZAS$a#5U^KWMmt zHtpkS*kVqqO%XO`FYHL1PRHCJE+sxp{wnT1nHc#Y<8~1Kl;fvN5u!%VfIThtlZd#R zWMLtx@+vFP)Ol1De`Gb7Rch!}DTUhR71QhI6qM%P_-UI6&C86RdWkCX|bG2cQ&FTJCeK4DgKK9-WJCao|x8Pw@~ z$gI^4!<$6s>1CiH&eK6zu1XE^d_l2V_ro&Buca;Sf4WQGBfUQh7AD39-nc_ziv;0x)`0#r{+rHyix|mIf!O`#9nrz38o_fH2Erk zgu#fQ$tt){7CjaBCbR27&wwUKf!b;*@gv~P~1XNyCF6-$O#x#o4(ob-WW8 z?SSKVIGq`p6r3^JrXT@7=7=M`q}C}U3tSV66)r^v)7RP;hB<}uZ)qz%BHt2SI%I1{ zt{Dg1iYKiMJGm&P<`4C5ppg4`E_GM(DP6L~3IVNjBr)8eUU`pbG!O6@B;hm}3?a32 z>cjXFNL1ojK82>(C$gjK+2AM zKfvT%T*STMEcp<~DzKMRux4e@(QrNj&Bzx6Yb#HHyPv@@TwUShqL`FFpuN#(W}u|& z$L)*3b=KKaeeV zf(($Nq%)IZlcR_CSI%Gw2XKOi*X2^llIy3%;z6tOy$f57zg4ySqj z&Z?i~R9NR~5tdUao$@KRzS7JfHL~t*6xxvdeaEBMT|l!pcNu3~%*I>1)GI`ma7@9c zz49q*QB#7_*h8=x6g0}0 zQq?^PW2~-la#4JcKP((dK?d(_|HIDKfdHkWVQ|xP(}+r^ti4+{@4huPSu--@&}4HtM%IBQI4b?NhdYa z9CeliyeEg_yi<}5IsB;$d(L7REGqljEt24VrEHgSQ5vs@P|MiqMBxnZh*2@(ST40V z975d}shX3E;%fZi+Rb{FFz-GhE9d)QL$Qzx^$|V%(k5^tLw-^>lrfwId@XB;*-lF% zUzQR{W$9#+<`$LM+d=| zvv_Ggno9e1s2_ee0c5z6GN&BjbTTNv5eT~Y>=Bd!u%L;W)!t3R_11cGmk5G>XfCH@KNbK1YvF{?}M z|2m%OrZx>A6Y@Wj_W^4B)tKT{f1r7@p4FfEX)_!e(16l>R6;yL_6tn47sgVCckT`S zPKK2t2r03jb9xesvM)^_Kb;fXxHZ5{Q4(_IOsyxhd0&RnUG;}Fa}OD*3t$Y-h;t&y zbqvdns_j6;EA9lOIT<`oxePICaS4#?^@IROlEl!fTBki4jKE_)00ypq|MPj@G`rf4 zp~Sn8eEV?Q^g}oN5YThiO(##;IA1zpRv>3nwZ7&@_<*psR)jrvmWB9(otvx8jHc>l zwS00k3A>2A43Jq@Kh+>Y$h3@(vrOW~f)4hgf%S;AQ=cB#Txp34nHrn-E`SLk0+A(j!vXy{{DCSfE0LV;=DdLW@jk6I76v?>@nU zL799Gpz#z6+RHcqpHl2AS(&98<@*D`TmBUx@LjUV8#NRAPQWkKt;YEn5H1A_kJELC zopAc@-Ud_LMd`!f3!Fhw3FE3|-vpJZcR!-ZuWW%X=L9MLoV>0^%~yiS$puL;Ju0Xp z6_|mi-C<%#&2|S9X*roL==Kn;0RVZ{GycE-k1K86qM?DHEe#7xRv@WKQnBKleEwGn z>Z8>tkd@$pewX^!XkIh+gY6NrcjrB|mPqh4;Q5(OxFunD0a7wjF#gN=G}oJ+I+Bfg zvnXcc4~A{kvv!t!9Nc~y+?Kb)Yi%L}Mnm(>=r)c_V@yB8R>{1P6c|ez=O~Pz+qljW zSgE>{Z${J}6IHfp|8hDnqu9iD5aZ8~Z+yd)a5(U=AU)6kr`xl`;_uv!rRghqXf%tG zJkyxbL&JwI z^}sfC<}@Eq3x`%$8iiOmWai@!KdscV3YSnG$25@O*qQY(8Z$AWHcUlGS+YyPRlW|R zN5UZHf-XiQQik(6?875=RR_Fs?P1l313=jy1jomf%%0MiF9XmxB{J_JBB-v3g_3Q~ zLGu#%Os83SLT2+qb&5aWv$C4Oi2D@obr?KfWT%P|uC^>SH-YAaccM;?Aa!sFmD@-w zxC=#grOLz=ZX5BcG(l$t-}v9gxYNr7Pv z%5+@bIG|&fD2TSt&q*EO$v*PLyZkiSsZiGXD1DSQyplB*WBZD(*}a+IT^?`w1=n;_ z8RNKC(`i&@qSQZ)0 zot{Y$O-<7Tr{#PS9?6u9-$B2Wm=*!aXau_nHYSUl5}$0x=&;c_5^9r2V-cu{b2ld5 z$mpaA2>`sOA`m&I9(7f`1feshxkr%E%o?v?A?$hT0pTZ3xFcp=3)-z0OHb3}|sbf<&xSN_Tn_ zzY%Q+8Vi{ls%@)Tl$CU4A_mivN{?^&lqg4VM)bNV^0DVdP5GP_^c1P!mnmdCs)sk% zGM@MKt&hn)E{d7;!KC|d|McId<*l2L?QRG;+y~QLY2zMzCqm;C!>0~_*Ud+>B$rqK zY&36EUg0xksZvqf+oKH{PR&Tk0oSR^g$*&jjrA2F96E(h`1s&y<{|a~>8-ym7xH#+ zp^PHujQ*N8C2>2s06wEtUh}9VzLuqQl#BqC_rk9<@+gzz2j8k?&`$;}xa%z@SYwuJ zEPNCWFP3LJ3`Jw*mOv7Rqu7i`#!fH9_hdKN60YHMc&&0r0$MgxhPD>=!Y}Z6sE{`|E()>gAb zR#fbjXHj`Q`o>m>*>yCzG^}v}H>9sBZm7a);G*@@{1ekvXU6KG;)-Nc1Qa;%krL4m z2d?IqIsFle0mbHn;L8Yt=PuJ(`u5$z=W~BpmD;ZHT?&Cz19o(U)Hh>PDlKIl5;wD> z-eVF95~sP?IXtrsKfsoqTo7;XuR~Hn?)7ns@Fp7@cxKyOEhZr|+3vaH>j`WRF zvAeg8iyx412Fe%RRIZ>dZJ)V1_Jmy08B^FVzvZJcnD=hA=QhT6d=s%=1mJ^J5>keO zo(W&`6%!9MUHIr}#Aith87Ix@Dq4O}EX<}=1Qbg?%427w(#pt369`mDc^Rx>a4w0b zB@p6Q@<|w4Nu{6=5qCQ%IE7lO#w4aV3Wvlr)uoru(Q<&i#^6b_8)H~2iZOU>vQ+CC zgHVU?_}E5%9-Ik*xG8OJog6z93aCL6hsFeA2!Ut%!6?BgQrfZteo=wp3JoHTl@{y~ z!CNR+InDO?Ys(6AD-EnYO<4OGsHB}-kf7qRSyF3cr7#QrW_f4May&X}Vr0OFw;Dj{ zr%OxInM^90=;Botpf(}*ARfjnoKJ5{hY&R3@3avMgv3vR6rzH&MJrv~-k2(wL`X*G zlA+)VpR(QedB#yK0;NZx!(acSQqS@h21vM0dHVV-kp-tZLmwjM@N7}NjLt+3P1y#L zY$pmIvJoYtG3a0kXvpWuIOQ#DW`x$=J_47w>WzI3G*!z)2R}uNmcfyH&=ZDI^TXb; zlMCWS{IxbJ^{gDjB;8CPmr0MID%YenZygS}e*_jASzAaY?K2dB>VzCj+O|^WL#1Nn ztktT0)vb$6q!WzhARmx(<|u*DpI4JbX$rZ?$HAB2CeaQo?WmXUrh4RunAwh)OkyS| zgE1D5D@{uWY)C^A^;-be_Wbq=2GV#!CHy=e)hGfnRX(cXPw$(5z6^D)jiW;rkSH^O zq8y!RfoQCF8fbHxtFkv38jBnHv8Z6m$e73OBhsxg?%hZwvHEQ>sC6wRZ@^Nkc4)h5 zEI}X5X%H=6Wpbrel%VBNzq``P5-xpnz($&+>UlH1phrIo%htZ1s^^w!zXEgJRYmqG z!8mF202xv5^%rtnfhIyd_i34^T>?aNyURVhpfRq+SfR#qDsT*EEayMp=gTS7)lFg$`@I+n%c zA~>spb|(ZK4MA>wjfQh2*eNr^v9O*3&x0MvTEr(*0WNR`3$niMJfUeaTJDU7fBls!k!ie{W#kVLBD4^ghHRWrslku)8^W|wW?vO-3;$ZQ6# zXQ)~UC9cPXO5bbOp4~lqM-i-eR=X-H=5r#|r|dg=-aS`~Knk&}#Uq%zR?7f$Kd`jf z0sqRBmyq4Ny8G9OB43%f@2HUnd}l1Qlq->&h{?*!9Okz97VTlgRVCK*Ye;GrfjeT>`6A!o6ysB)s^txWYO9QZ>gJTyBb}J<4XM#eI{AEWGJE1 z8-d2yU#_`*ZoJ{El;>m@fqp1`+HA=mL|m<8rO$Wr?1&w4t8wwZ)FnVOASzYYdim(6WFwINxzC?a+B%pS!N)Z*)i04%T?XTBNW%=71Bwbds?(r1B1 zoLrES%%gm9wY8EJINo%Qnq*_7Uy|Pud$cPh6=sm}`fl#o_CC0b$s-fy9bU zM98Sn^A_fF(%@V7Gz>XGvoQR~tje<-JBAb0^maK;%xcUKY7pb4iP);SOk+XED z*L)$%oLm%Z@Q4PsRx{KTm`2%%m`Jpt6sNi&+pwbLZa7u5h;9`A8S^ST*oCE|7I*Bc z)48AlSMHy}FUMAd0cfy@_syI%qG$z-b-i->lDaP}xn3`VsFQ^okD%ddRyXqwOFl;6 zz_q;N?gQKGx~|u?F+}~K)P|OonoH}TV?o} zFel{?AZ6iXMrr+%wYD7L9ilN6Q)oWwIC9$;U>8akw+v$ygJ>uEQqV#}lwOW^*2|F0 z(50j%`^ZGqIX}Ljn5Cu5X?=ae^ElWP??h8Hwj9BEG*Kz6!kmWESb4%y#+L!KJqjAv zDzyw$?09<*p|G6*VSYV$$E+@J1997f5RiL4(7rkwpCzbFV>K{~BCybZGkh46)T$LA z@N1<>2BHaC^*T)&dVyAVur=eU__BDJSS(R+<3|i_OmKw zPA-a3`NL#uwMGV4?_{$qeOr-@ugaU+^K5dr2_C=oZN{cb8>2FOb~lO#Wt>8tw0&li zxqTpISn@Nb4|g1}w`m7GD3S^7u;&lB5flBo%Q~N&LsN~XjA_46nDED>tTn0`m8&C8 zUIgc~s8dsIq<8WF<%3AN8CuS*u}9Sw@&ZU)^H?&Tb`K$YSZFYkWZ{7bKyXmO3ej6d zY5+Z)#+tqnfmgU{GWtZ_tUw{gl5+}S zeXV2Sf#rc_IFi}>1oMg`$zp8449HpDLY8=W0_8YlO%bPLaKE(x!{vr3`6Q$%cHP{t zq*RSul0?+8lSw+HuB`#61SMzQvFd9$sj*s^5aAC@tu?cBiDR|eq>GOuik-?mkp^*j ztq8&WAOPly)4TpoEc~@Z4Nt`jnlx5}!ukiqn4|U+C1mio_^l*`b@&?AlFwM0E4~Jo zIJqd!#vj(oI^V2_bz(-4)SEFYJROCcb?k>Eim`IgvhAcS1FykMNUY~`N9o2AX!8K7 zDn?;F3m(a7)Cbo%2rdzeAe4n;?S?~+=qvCAbK57;D7u+n4oZiY^RcFxh+`D3C6Tz^ z%_lSDNanL!6G{f2h4zWS*MMzBDk4X`8+3?yZ9n!x4HLstmuG{ zMh!E8=UrN#1h)g^AhyO61oTn7jmT$&qH0|TL<4q|BEN1FgeY-WIgO^_3uitnl#RIc$pRq=80qN{Voa&GfQjB>*nWeI70LeV=+MEW8Pnv)<3b9ao?PGAE%ZLm) z&8y{$``Bz1WkAlOYH}Tg$H&#qn3Ww}%>dHxX1O1T7cF$O9dGu-*opj$8wG-wj!VW% z0iqkn7T}u9#$pkl`48ao5F%sx(>Fe+Itydl4F+cd4rvcL$K;4ocq!7S?f`@3htyl# z&x!s*{`+nI^Yh}{TD)A_i0;$*6>GV)V@VGz(x5cG0@TBe`~EZJ_uu=Yz}NjrD`QI; z=&valu{6X%ZulnwLNpEfi8d;zkC2~#Pk^)5GPdOA7x<@)?m!-?18BViK1j{wmq~(K zE53RIoLrD3*rQ%^y^>`eM-ftJgHgzK>T?DlZB9Q~YMRPwHmy`*p`oY+w!2izvPD` zl8J~&rbSihtu*-pR;%g@>fz*~n2|rcxn9XyV79q!RQ*_ZWGys`0L@-Ygug9BS{@j?8&w5 z=qCgMV}IGm(OG(!>54%Iw+?g6x-lUd+Wlc<9&#%gteA%KdruKx& z<7FAs8wDb z`^mB_-d#d?C=BV`9oIcl2 zJ6GmYgm$hudw<=y`f4i!=Un3PsoB36a?w=7o#t)A-}rY|gmY7IGnETlnu5Kk`6kwS zPJ3AOf?>)YRttb(nmzbyxYgHM8Mwf4*ghA&lx)`FJM}#l>`j-eABT<`3i9H`_$ zl+guSY_~vv4_10_?N^H~wBi=Jr4B7J5Yk-7%jl+apA|pbx!lPG2|WH9WsOQLV<;+- zlUywJkf~^6UqHW}(OGQ@eOy@dy|f>D@n&FwM%_ST3AQDe@dm(1b;+b|K~q*FLQ#U9 zT#5yqxXgP?v&0pAdm-v>T~<<~o(?VAc!-ctEoI?;m)%PKTa-)0r_jO zH7Z%7{IQ`J7|@=Ze555XuHbXM5;=$zT9Cq@PN5KzU(jHkDY#4rR0B@cox^eU{toD6Bb9lx?6C;Sb`tt3J>Cqa+S8FPwug3QD0X29|mH#2A9$9=gUTsbH5 zB(7P?p=gA`x~tMtN+2YLF9QsVuyi^ft7D3U5MQH!Oret#SY%C&j~{NZCGT zG5&g7jb_#^xepuV!%v5ZM0^-uj^Uy^PlLsXbI8`0(x+&`bWFNbPT7NKr&Pl>ZF?$J zE-qe$E@}|UE!5KDZ4edrk-dSK&hHFSp^@R}t2teYpC+gbqzFw=XCVHNP7EYtTH!dj z5hC(h@y4=@rD;fwbCXtFsZnvhG?$=3NTY=*5;@bs*U*_q9H1Mr(Z!k>oGcaE#ZFoW zZ>KSrea&(I<-aOzJ-M0QUa|#saL%5OcOl#F%I~7f!Hj$cPBQ_Il(sVqFK|gmC}w}v zGKU#mjzL7CJYm!>&x9P7P+jptmv<)@B;oO>UfpPAIia1Cjo_Hsyh*?wmnHnoy6bTY zd>od4rCh?gu=4cmunHN2&K1EjUStSi87v;A7p|RX{1{(WF6T~@$Sebf>^}a{N z$|BQqgd$7ZLx%Yi3E5Qq*&2o4|GBg^`Tox=20|8g>yjzBQ<%k5R@}lPBpX&NGIAKP zN<%mCP^QFZj4Ecq=;&YnWy~MQVAl744%q2IstmYq(|UVgeJ~JD|Mg#nqu`d{g_rsM z--4Urzy1q6#9;*E8DMY-ALiNXUgQY;XyUn(i(+*CFxIXYVN`NURt?fuieu>x$ehR+8^MGPL1-3*@HmJ`q&PxsEi!^|A+DLqs|O$7$PIo(fzV7W}IZ2g3HoYe-@&0G-91c_+^aE z;d^hVNPGP=7ljPt6VDpW6b#? zf0%G{rIF!_wl)sSC#1`FmI3i+N#b={LYBt34MGJsqLf$-b2>)iO{Gxk9XO3X>n{N` zYD!DGS?OBdXeh3%hn+%E|9A&n;?a`&vqmP7k1uTq^=(eOLgB*`6vW9zNs;{F&do{| zc=$1zg|12$?z};G3Ma2iO}a|k4=R5yT?#gvF=bv+@Hpr~2d6HNH;Z$e)nNBj{grhq z;dVICe$*!lYBO+GIrN}}Z2TzTYb*N0Ar)a5j;B|7$!N8j7lu5ph?5IqG9IPLo0TlK z)uJH*T9|MV(PLZM8RL?mojEcTh{rqS6h6F1AWAtP*0fF$sl{l>oe>xgfUsv&RqbK; z4FeISS-CV$=~8~lsiU;V{n9~?L3ctKNEAu-JkJVo-KZZ?KX-CbY{eha+^jS+_TNE| zFupRxiR&ntLFRgCTCImOd`)Sl3p}LMSKjkB8Zbv#-85DukU=LLA(wQw{3>0*rL2iQ zL?5a=olz!H)!7PZN${;Hl*YXCnfiTi2&J82N(Y)6;6G+A*N@OdsD)BOe{0}KJc+bK*fI-*s?tK5#gaA}( z_aZocEcsP}@+JDqAA{22U;Y>}t+S^Ibt$?G%ShyZ`~IhV-lFb}g2l;2@fQBz=4LgE zgVp)8ZSp7-+isoeC#oZ$OK#u5{^8GKonsibP# zHm*p%*j6lTL_;zebx9NVlyW&}n8)mKFk*V8Qe1)02VHA1*Dj>k)+dTg@sFtg7w$`q0&smJ95%XlXx5R>%a@R+~Qp$;J!hIM*|C?X-qpz?f z$W-Vzl|D)%Wj2j8NNl^E(Ucy~DRgpDed{e>_!}n|#eaCjb?aFhR2{zG9<$XMi@2L6 z09)%s!>?)xXnbi7sRr7O)ZL=;Vg}*upf?S!84bO?rRSKk5R~gpKhiP5Vdohx>%K#~ z0)Yj%V4e%H#p+FN8FLjO3oB_&5ItFywq9&^th}eQ1yu1*0Ty?V|kX_ zJ&>~*Vg&aBisFvZt(;ER$kfP}a^IAx?9XW%fx!~10$5hN!8n3Q3b;EZ@_@|Juwt&Y z^bV*-&7(e_1b6G7i4tM29cxB4Ii!uGX-vN;@6rHuTG;ulsIl^d$>L}6ayYpl{>P(0 zdULIjG1lZH2;^p#O(oVENR~{xl*U$W+~qaOVtzYL+3YMmMs|Q=2n#RczQ)O%&fV5l z{EWV#YSJW&a9v~I0OgSZ?~)6nhM&`foKmz-5=!_S6LatBT>ir4oin9yQ_;p1R>ys+jz zg#bg)PK>4MG)4VRWeplkjdl3wqA+B*OPV1!K{Oe=g*Wfs*jJ^ zgywxc=n(;qqOuX|w80B>1SiuYD%j^k3f>TDS$-D1hY&xcTsEfJ~gVg6TKT~qW}fUr@@df&`spL zS?MFUxr<5-_?0f!v1|8n8F24;nb}irbO1unUQkxQOT4oaPS}^R&N$h)CoQ}-P2TXD zQ?7#vndaeOidF3gIGTv{)H!c=&5H_}+{ao0RM4#SsQE)hF{=#F>+sD!!UzB9KY@!* zqFFS%xv~&(Ao-hQwAsvT*}lqcuJ}$3-~0?#!6S}QD9oe`koBRxK6MCb2X-PrOqFWs z5U4{{YyXDXW(a`EYVkqK`k!w3RvEd^n$p z9V;P?$>0o`6fW{jS8!=Ev-p~uOj21gloJM*OP|O?`lUBe8cPQ(+~~|Fpq%%3M#pn& zu;6r9djj`=De$NaaXmgGvQb@NQn;+#%n@1#zfDYgp$Os+v24{U8EyAud4+=$wxMPv zu_I>uxOvjw+t-1IG-vqC|z~f)G)XaATfyqbm%Qv zkTqZ3`-yHw*bNm~qYzPU}3Z*q~^6t7DX3~^35x*lRKuC*?yHLyp5{>fBLQkKh8ni^1XYi*Gv-f*yR z47KB>^41}Z&>TCb+8^q=C$Lc24XJb@r#-Cs0=1iKMFF)vYIwIAS*xUDGfZp1Ki(pF zXrF0x3rZOqA+ac21v-KGcE`Vx07~ zFb7LTki(>19cnJSjisOT<&xR=4WW5AxVm8DWJvUcQZk&L<=G2JxN22=DRgjhK}sZ# zI^(M=YZ<`JPYYk#zPg%30hN#Z<2qwweOt4m&{wVe4pqK#vcNh?%oi>-y-*lMO8lAg zd~$f^>ZIJ^6it^?axw0WI_*5(p2Kt9H=k|MMm!|6ipWa-!v17OgdPo!|wc?iy{f) z;T}0N(a4OYOP#BMZ6uD7`&f509jQ)j)jZ4(^-_cJ6c?2!{yEIbiHQMC~~W*-t(_Kdt9y)P8IXkCYLAwde2u& zC;C>t*FAr+?6X){DqdW+CTh4aETP-u9ztsHIC<7yXlNF1+VVravXcwqr#-emvNqt| zheRcUh1^Vl>W7+MJ*L`luc?qOJ>P2*u2Xd~2540w^2$w%hUZF5MQS{+^8k8@Cw$?O zMfi2*B}Gqj&V2q{)+HC7^TA*BQ-oq9t8o$fL%F{4IR2U(t1B5=4|j9=nr%D^2;;_| zT8;w_Q~yj&_HY06-=}e>wlj(%s?A4CDENY>-@{7?;jmq;3KPV1$Man0XHfTzwC+VO zVrdu%!zuQvegeEhBPeEl`KGHYh)4UW{Osg{_!fWtmP)30^wDIFz0mE{hB1Kh(pgA0 zxX2$$Z)Oi&3$4$~c;vOM_B@+7b7uGzKTC95+-?z;=vo@$uPsuk*E76=fz=Ervw?e9 z`f$}HHyrowmKJVAqtSev=(gUukmQpNI31TakwH9{N-zXH5P@_b6l4+x4%x(eVhmMh zT&nNkSEO^+hvt)OK!jWLqylXx7v$;p*Z-g~o{^^>1v8Ky;7evZjH}}~6UtcK_LAjn z)U4sR(MwPRZ6#5x*c;g9(|t9~$WI*_K=ok^3dIf8f=+FoWMn5NtW#+`Aw;}PiC6Kb zyIg5CGiC#S+N-^W4|&ehM3HA>=g2yBVM(cKXCPzhgi@3iQqJ3T*`}!D(AK>H0}+w6 zI_ji9(fqPxpf(Q<)GD#sBaPp^RbI#(m{>1#!J}xI19K9D;=zda!gcb^r&o|D@;>LS zCm?grpxyJFw~C;lqU5mR4->3r$~kprlQ08$0vXp?-q;JKFYW`2pD+*F1(JC-x@Up!lzJn+Wfj&_qXq zZ%zf}6rLPoVwjqnCfx#qfIH_h#(aiR^)!{kW|~4w<(PZ$2Om~5wUIkVWm_}YLH?F_ z0_DN21mzMpZ^k%|^Sk7acHZndnp~#+guq(lZym^M1LZ0URSg0n3OneK9cr>B?&^>S zTmwk$TM7WaV4oHwOxlpsOsak+Jus7M5r~v#CjL;(TBDXVyhW_|id(Rqla4>J`1Z}1 zc7j3A>uO!gZ7hH*Q)P#1F~5LtNzOvrL57(sEI|%p)6l+$$=s=4bmqOmWPnAY)iooJ ze245UTknu=3T^`USmNr^6TFtcF>}jlI(0w77MM=G2ooS?I{q-s8btNS33~g3Z(+C0 zG}yVL=iKir-3u+IabdP24L5PH%AW@8DQ0C&1gY;~{b{{r2Ryx_@;EfYj*zOv*8qXy zX>6AWjtO|vuqt_Aq}ud_e{*tC@)m#CX04Sqw7$gvARtpnf-=ksKSr8i zvSV|O907%n_6k{BE9wts)RtFrIC-*CqUUcjub)+$aI(AI9}Zl5lqs&YvX*~rb$j8* z!O8SB*wN@J6hy_%_4>0i3?_anZ5$j-OK=$WBSuSsCJl{>_j~*{9zl_6M;hHbUPZo6 z_RE&C?YWvHcn)LW0g@>RAxqgqpl^DMDRplL0rMWa%4o?~^!H*GNt3Gs^dnDh0Zt|? z0b!?l7)yK1w$)imXH+Lo7w8{uGT%4p&HV2wU*Nl*LTH~6BJQTeAL9a}rust=pxOebO`s{rMlc|R& zerT~?ECjfAnbrKCnhI}ZnB?d-J`wvinjJ?Bpgyo=BAwyy4%ndOkhC{ENPsHUj zAUT+tHhy4RM>@u=j3zx+w)g$dIZdhVs~=yh7l9P6NK^F(xz?*|8O8FV`b9A|PGoH7 z0;?}&E(xx(ML~O0-ugDMBqak|IUU@Z=TN3%d=8T-`RvyB?rF|IydD)fuAl*N`I1mK z$bXDkYvyCP%z{{_AHrf~eVEjXI3`cKVJa)NLhObyYw`!()>~P^*6nTp;o`N5;0X+V zoB-bjNYSRMBFFLI)2N_B1B^d}5=%N;p*j}7vlWm zbTUuTJ$P0+TVr@^I={MdrJ24O+TscQ)6fKCC3?WNtl;N=)j`$|6+cZe z23q7y>+cF6re`A{+W}GwdiOZI4RAScnGl)af{=p*69nhQfK+Y^m>Uhc_A#Uz!dpHX z-GbL`O~8{oa1W4p2)YK~jYzqU29xO^_zJ2=L-G5cp|&ZX%D3zczu{@<$xbec>G=bL z*#exQ{}jixZ7pZ%WuY>$vXiN9#Xt->QAsoe*|)b& z0}@`BuP}`d=udh}?||VG%Sd>Q8G(Fpr(W|TBMKX_RsBL{qN5Gjr*hgs;|a+83#9@6 zSe%V|E2FzI{I~!0|0ugI{a8F8_2w8yW&lh(w}&AyRZjYj;n(iMA)l(1H(bzvQg^i* zbg4&QA$|i^8l3SRZabKu(X=(;`L>$FWc$py?}g`VP0o2so_3k=v=(BQi9N0UK+i^f zErU7czV!7JLAD=&R0deNH{ zm=pT4cN3=%FtK}cQ1Ljt=?E{Bmr{mx5U^{#ba)wz(D^d;1PYJw7JK^JBWM~Q~cmw~;T&k;4AG9dh{a+Q+{VoV;@e;ds#3hQn-yt@jJ#*>qWZj3Sq)VP|UEpE${peddt>kU9Z6 z!cYdnz5+z7YJ%3&7u6o9rCdwm4=qs?D)^o@jOo#%KtBWJn&c!PHX6QKw47WNyYk0V zZ9WByDjY6o9HQ|9) zA0)|JQwhZB{MNt*Yv4=uKpNsO2jg7;4#L_SsI?EoDg9uo7yXyh7M`-B@r5!Be}EyY z5;NRCb6->3@9Z{7;WS8a97>RpNA=Oh)zdD9ABB}A7w-Ftn%+tPPA-ZI_lF!dtIdqI zc;^VNAA9A;fQQNewiR+GkHE@P#{(uoy(3+8qKg@FF=}-^?!u*Q0>aGh7=sU4o_NCp zEB$7p_^~RFKT`9EWU4lk=BWAQ##CRl?nQB_Gq;bni;7~KWmaExQA zaG><()*Im&1Y|0K&F3MSPpLtk6)iZqDCXx6DaHIUHW8QwNiIjWCmzOz)K>@i2^i9plX8t;9GG>SGL!AqTn`n$eu_DQkCb29vLR{WbzH zX%Cs2$jcCI)}H`rKhHwyMQN$m{DGy->S|U#5sk|vjM&Bqb;aWj{!n;$oX^pJ&{zt(`YB zG=`CwkBb+brzQ>~m99R{a7V-;CZnX^LA?@0SE`IP`NAk3sJshgi-8#Fu*rQbkW#N^{Rd<(fv9XffsHdnS@qs> zg46i^t-&IrP{xyQK@!)q=4(EmlM9lTcvKK>)~gvjt^||)FeDAp$br%aCBE$7d{l_h zVo*H%B1h2)|1dYUI{jSyP_jvHv5D*Pbh!hmOW;W#3fB!AFxX>p-R;jOH@~CW1$##w z9c?I0U23Q6C4=C(1bizf#Q7qwbFwp_A3j9oQNXuZZ)6~2i2*Irk^~mHm$=xD&Mlkb z9bzr0%}e}?k|ve|2q!5o@(%QNDEMm_b;+48NnJh;p#m`Xgt5pnVQk=ih$`)yZZNqd z@qSfohr!n=-8-r;Ii;yF;^_+Hi$$%fFV+_)7sVxc#9HfFLc)s$ z^Nt^2S+GHBz@hRu$K2;P2k5I)lr|;zjw-D5=o!o6>Pn^zW;Ym@!Rd!?H~l@S5qdzBRbS``pGLXl>A1;?pF+w`E{KizYZEpbwG7%C zWeZ>k%0%58_yEB4Z!~$8$QYZ#75ks=4`40}R+Y9@9Z8I&^npEoATcwIPT79yV~Xl7 zucZuT3Cxpul()n_qA+(VWKQ|CJq*v80XDE7MH3~@r{q6qo2WN%A7KRXyFmc`7xf12 zp^f)&LFt^`L62?)v4anFoF+tBCc@w0h44e{s^({Y*iJ5p#d-{i&1wee{WzqRt&}{} z$~XogWOaJ6{E_yyQfa;SObPY-qUiumcY5j+Zxr-SD1BK#T_ydJwED}crgPa(X=LYX zR3$OWW9zV4%Sf8H!|vV*L zFVP-gy~@0&-_^D;Txwre)DDePmx0zHQp4?q*}Dv{`M?;%Gu|l#*Se0SIhiuDRx~sz`ME z0SE2mg7_AHEz1`2M~@R+4#(y}+DuiV((7cLT~xw^!U;htZV|*hCp-@2xD-CiZ^lla zP2x$WXp*w0xAUnuPnz=ul);#s!mH7roJpxig}gJ7D?+uPJH&SIKK4Vq*q1XzM$AXK ziML#{ClzozxgZAQucO(jt!5a|Ui+X;x*7w@ZIiD^;KAjwkCbs71z(2|+{mE@vIOeA zWdcxz+K`k)<5}P4&XZs`LWpM!4j&B8DRkiBPcf4=>?{0F&_{PXUvNUc-V0F|%c%Wm zkS_^N{u-IB##%-r^LX>H3`(>04SC>ENGtN*z(&l3aOll4YG=$GS_7f&)P_W>nV~}z z2S$%CuA#_Xkh-bo1seZC(X3iLCf|}(@4q6eIiq9R9@2jJIQSCWSZU5)!f6PuoHO&S zG`yKmqbMfiQI@>5x{_f-AIj?y9GcQWyTh>I4C^h_>Rb={m-#Nvm%Mnj0P^CCXWk!b+Qkr7d{dJRFjEc0)jt z4qz*Q4Q4t&o3lb&)$b;ovb$u*!y)6TInBkB#5i8&hf9nv%*7uDTdCDDrcU+)BX#f6 zPQW(Bn}KDDowR+5NRN6UIfrEgJD`93!W||&PPBp z(g;nzViNty{3{gjeqb|&*v8-rz7!uopAD`@cq)m6A`}dNKn)vBzXI0}5T4cy6VO_0E`oKnl7PzhfvwE( zJP>(bl}=x|9K}!(b{K~;IqH{>gG=n8^zjw;f{4L&Frwl;o&(lyTZ(fy>;hpCB)bs;kF#2q&G&w)is(3V8q@E% z9OmipVR@Q|wTd70o$3lF7sVv}0kkZD$Y3ZOFlK1uj(Z`KkLw}5IwZX;_>d2~s8nv- z1pJ#(!2H>LbsaMnU{>1-AAxJ!8xdgfw;F$>xZwG0fRz70^jv!zNuNGN!;73&;Yp<& ztN7t6={~u-RK&Y zw?Lco@RB6_EW)DJiBu6QH9oTocJP<~g@baMR^1nxU#rJx{tufYSeTYSaiDtU^y5!P z@gAJbEwfd^%W%jQ<<5!Je>Bdj__INiZ=xh7&W5#=zBguKBpRoeh=S1wcij=}D*V-C zKEYa;m>_}x5h9YrqVIn~dU9P#KO7$t zo$xuA!T0~0D&0VP))q5k%|^-Ze`1bWbJ#QT;Bd}gCvV@Uw(6_z!^s87WIQUBH|kl% zQk&VWnr%i`W*)=n{EfWZNYGW*-TzEA|BrQ>l$w zfd3JluaM!hU%_8?prU}T><0k9li&-c(zzLGMCoW4er2bv>lIG|34ws;p6Bcu`H~&4 zuKak}@rBX&!*g5ptlDQoIU@z?@>QEk+`U2vqewXkf|K;e-$g761%{oUgZDE!-J9R2 z;Y-m&p88eu8gC;$wvJvQ+3L<7uLt$yxH@8B)3a+kRz;(5p#Pe8L?=(7B=M=MzaE3kaO1}v>&^XfF5*y5ON3NvO_ z78$!s)T|+Ig6{E;&>mwkBT^NHRoZwtM0R>AK#gG9UO3GQN#lO~~bAo;;>qG4%vaAs1j!McPI)me$Z;3d5Vbzab zqn%t71MvrG*0L5^b)xy?JS=asxjS|yCF2oU!tY#BVnY&cbpD3cd<<7P_&Xe4lub*= z8J6M?79#k;nl9$y+B9OyovSFiz_FxE=5&58e`N~vryx{uU71J|YoXdceh8eL`7A<> z1c=KgDtKdeCi=4`DSqLP%t;?Yje z@l1?Ex7|Kp1Zz}%^;|f)D0bry0$r_SM6kngDARb&Vw;d+lf0=P4ZG%7+FC|-uZ2Qd zqlCqI5>QQ)(KSs2h809`NkJrlCIpyc2)wEsC$R&q2!_aHYd+C9f5#Ac&Q!MX*Q+~f z{=mmdC2J6BFE|T_Rr^&u(mnnG=jwSaMC$M4091syttagnSRrtBrUoz%DCs7*V)Vdo zi&Rq@HD4@c-R4qXzVK*x8%W^fqCCI;*rt_QJ)>vZ?uCXh?z9h*Y(Q&{cv&=L$0lFO zHmHPVtRo>mT^ha>Ng0Sa4K3`1VYw4dPyn6sG#}3gWyTX&>nk*Yr{ARrE-swT%SsZk ztDK-@!;{!vb%m3QVkiDU$x1Ve$G$tDEiUOrh!sLCW{Ft%oAG%<52WHI&DnEuQ66W1sARQT$q*Ad2?L&Go}f=G zBMVD3l=fgYktm?UN+~F~QLH$YdqB^-+aK{@XO+YkkHcvUUz)Kjqh+ZlT_QsLhyp;F z-dhIYgjn1BG(*ug!eZZ?BKO;4mRhOdyPGIF160eY?^|7IR(TWE!yJf$vpp3A8frLE~vD!f{2Wk3vzP!kU{$yw;y znF6420A=5X?!+*HjAc8#E)(-vUl0rza|(1Ok?4>HNZPHC4A@#lyvqFo1>3X5il_6C zc;%L1ymU_ld2de)--U4a4`)|-6b^^Cmtio_WoURgctQG^7M8?bxd%!u_hy#EQ_4o} zVGobv4Vr>eQn@q7xh#VZFMu{QTTkFvy-cF%QH#5}nzaYODuQS@w4RtqZ}Fxn-QL}E zYtJ^!emp$&gYCDm4=mkb>e_@RxO5VU!1I2hUD+NFh7h7m8-aw=kfEFl?(qC;-ZF%> zqGSjj^}1OAkd-0CI~3y*Bz@&cA7e!L2OpfOli{RJQ^vq}#?4Z%(`L{YrH3mwoYsm{ zOU~QWYE_=l^?adc)*nJxuV(3)B(r`W#%BW~8Mw4fPSij^RRN2jZBf|;Y_tf%(iS77 zBj=KxnkbLtVj*EiTdoZ$fL=sKrxjrLkYGxF(Gjdmt$jmKP~yk@{Xle-xDPNpvE5_A7<_JMwK; zgh@};t#q=5DfY7K`blq(52+)3*V+RG{zf&goh!y)xkqQw(p&O0w`EL+=6UG36$%Vlgn2-9sLT+mfq}->Vs_IKWbxv^IUySw*~qW-M4Dh%6e(n{Ofv$z z%qwJp4b=RUeRgs|tjl94Yt%E4NtsDizx;py}~v#QKrmm<@8BU zAy7WgCpBU~`-g%B3Tgh@Q;kL|!)lUT_SjNXG%FTKZ3p~FTHyCmR@=dG4!fQ@e;SO= z2K2@s=}ch~N>g1+8#9bb>&#JDh65cM0DZ)y7t34aTA(e2)TFOSrGj&))E@`utTcp3 z-#(j!UxS-mohqDyCy6yXxghSwUx%!*R>|mpuQN1-uJ4W9i0N%H|BsirBS#pjEySAO zBT5CODSSF+rzWnM>F`Ry9N8!a4G#q`ky?HsZQjA%B(}iaqlK25|)}#j`OBTAD$~k2f__MQ6Lq<2Z(?4U{PEw6@*4c!0|%ZGekp zVWKU;qwCU6FgXind~$0%8L%%5tpQSGP0;NiU)~#_sCaTqlWY=QMJ%ZRlzzS_Kn*=; zU&$&b7v%Z&*PUWdZw5h4ZB4v3YB8Szajfm>P1vxy^G!JabiQ{2_LeSz8Zx!xC_p{J z!bGAokus?_(;zrUS!9Z^(S*Xq$w)9$GsbSDy@w2+X7whyhGejgr{EXQN*g%2AlBiJ zf8VTRL16paZy3zE8=)f#$e%2PX+RtYSMnDRSu0vl>jc0La)$Ow-4DqDv#A$>Z-i#1{%X++2s5$e+=A6Bx=tS zlxOHaa~#jQ4OJ5~-bVegY2YcIe6jFFp_+*Bk|q);PzwLTOx?f!H@ZyN7b(T30b+W1 z4;Cm3G*RSHYvUw3W2Gg35+l2BzVZeeo4yQ&Uuc8dPoG31Ak=9eAsYhNTuL~a4ikYR zCv;M2`e}t=lFcHtLJWCZ`RVtYhi~N3|B=`|`1hFI-lWsN)fBOC3 zolST*C_!+(`1>nR(#iDa@cjJie?~B@`m=k%^k0AO&ql+a*`bp`$l79%2VWASwT^WA zae(10Q|cZA-Qyqpl_YZ$-B*S(wG^8rZ!WW8`=Dyu-*k`NO%#f;b@gT;_7k)vl*AG`+l2Dww!JYR|bT`qFPGJOwJ0{CAa_)gQ zQCq3jd?gEBcKWmAm)wOWwgrugWOW2Jk8Ray|&@=}AQ_ zVyhMT;3TWSGAbCr>GMeFgwTk1J~IT`z8pu_ z09j&f#Csk#Lvc9dAlgItabpm(WQ$eWFf?xASXI*Yp3ypEO;Hr}P>xc|!?i{7P1oTQ zT2EN(@T{DXlZz4}{6V9YmDP-W6891ZnFX3*Om5%l2FNNT2X;WMI{kr^=h1;Q1gGm#Spfn5089G&Nx z%xY0EE}zQFwY3ZgGzlYH7chOm2xhdyaVo6+5!OW$-?cnQ4j+*mOJ$fE@UcDy_{Kt9 z43Bb=cobO+pai@iWgr*W$u7u->?J2Zv$j%Oc|ym1QQ-d6pUtkOuYdj-<=)FzQ!!e4 z;pEkMH2RrV-4#yq=gP{f3XQd&*IKLPS{e1+vLu|WCh!MFRvNX8nqdFUKHBP#6lfbV zb;DpC6Zj(G+4w~8#l5@(lk19}1U25d6aOF$Xur`T$mf89&eP&RM-x_t&qqj(3bo>9 zVwK5tJ>ei44EBu}{Nbw`UTMR!mADhA7k$xug+i_Rgd2O-BhAT0i30wh&1xfSgKlRw z@3OIlJwEdGxMjw7rwf*!H2L=cIF%sM5Kr5~M_f6$V`=x5G~MP1fANGKUs!HCVo2}W-fCi{`8!C42bhlrXwl1LiD5vOsq3HqHBDf8% zv8YYJPX31T!4Tb%T=M+*0B=}1Cl|z`JgOvD*XkL=7C)V!SPYQO@IOm%CxeR{^oT5> zyw@?);P*rWd2D`&P#nCPke^}`{OmQW4H@iXf-9|R+eTi09On_@Z4nB@n4my4MHrg~ zm$0p)=B!WX%PJ$y>wLUn*vR!1#Z>%Bez0LTBcE6R6Ij3oZU%2apTZ6Gb%N0qmz|an zYTwLRZc1`_r7mgv3_A(FGQgoEQ{PSds9&?CtOu3mBmG#uAykccmz^J4{C*rDlbvg+ zCBCEegaq?t@&})C!?jw*Biz;w7Brp)XXu<83yH0g-{@SOUv?udiP>JhTto?Izpxs{ zSu%A{rkD;kLWJmW(xd*JF`;51I`TX*>)POoyK05y}6o!Dr^UX$PafpCP*r+oZ~&>uND9|t@x^kMF= zPDoV=F7@Z<(eoh3)B^0YAauPcCcU+K9YbUQJfDTsD^D1zdsceM$wl$-{(wh>%QJ9j zc?wq)+BKJvfTc^l4eJR)l9_GGqeC5IcHa zm<=w;L2u_H8eyaajCs?Y=(2DD;tl(H^0u$)H9x9fwG~cw2J*v^>8d|qvQb^jST<-4 z2r1Q4oqhsqBvc#L^jg`Wsk~q$$*_@YUJ83^_ zgk(LkSpZQdwQ|KN$K?+)b#Kr$Cl|#+{Na?%dLz?9R2{fpHWV5}fXEhAm~I>#l)ln- zNW`T;x82x?;*ilG!fATXiY=yuXbsS)HB7*ceOn+%1ZdY_7F3AjnNFN1_X5x&dVu z#_zfGj>fZR8qj_-Hnh<+8OR1ssLv3M8?)P6egPiWaQF#JZ@T}5;lM7-F9(9IW zjd~V%M7Br=3bAqtgN!1W`Eu}Tal9Er!W?v2C=B_{!oz*Bx^yFZKN>NDH6+ixVJ-t@ zDA+AH{Np(UDB+9KAs7?9-MMxQi*36fe2Z6o@^aux4rM4bzUX*{U-QL=M$H#1ij#|C zGXC(!Rx67u2O|;?kRroMh7$ztwV{y&fZ2>dV53>jI&yMWn<~jnq{IkRN6|Ieh6+~R zha!1z4?3Mnw{i9%zkj1Y(b{wj3Ys7Pd`0!wnQs6+_a?X+Di7fEh27 zT6okW<`y%E3woeCvzsAPc6VeIdx73nHQs-w_tm8wkHH)E`BzVdFVfHcMmXaBw=Mvc zu#U?>eJ9ik1ER-7JASurhTs1@1hgD}|L1@G7uZi~ZBjfCFvt^;HxF{6lZ~b?&J!mW zLC*R(Z*i6N8p8b_soXg{S@ZW4rg0Pl&+bmkNK zoOBsDBA`=w%BV!?T|bzf1wA^R?FsWc?77lgAL&~u(>zeTV66sxSv~V2&FY0O=$_yW z`r}|Ur^%JGhuK(tLTULzPr5(!Fmus&=dkl1zOu(p+dyxXa8-&nmtf0iK z*M4x7vJY!9n_>uk2sWvfa|cVFn9kyFCch;Nf^-d-BJU~W)1Tg%1XL9~{BPBL zEbYJIZ_o#KBr-|dIH%=QeNiKu)uK=%YyKd|N~@C5wMZr>L372k8e{OlB z{0=UvC%q8-0~TcAWXbeQu{>o?$j(5AG=ZW!%erSFISr%ctLALh;+pe^cPncCAjisT z7H4*!rFojMdn1Yzt#JTmd%>Ws^RTN@ViBMowS;u~*%@?!AE;y3S&^Uw0_w^Px6+=l zS&8pneSC%QxE2tR(m!T#EhFhK7h6JX8-Ut;P7o&c2Fj1h#RJe zVJINyE)p(s4nzlyXR^~fdR9AZ^DWQkTeldKhVT3qj>{Lk%5q_y#Zj@ zr)YSQ^Y+)gWvFXK$xuCN9kVhgL!BnxW5T1aG$QV2sb1|Ep-R)c*4Wxs*K--fNO=#m z5JhvtL!3})GF$=Q3NPUHx8`QclO}l@r{Ls*Jp3Njj4O?1Mvp9A+5%-y(2}^1Ra>^{ z*50(o)EXe#S?h1!GLKc>zyjR$?}p`50s>-)A!JC&ke8)ckK3Rg+2{|VIl6ZMWTovn z@1NZ8{@%Q=Sd!0Ow7hZkIk_nI;&0Tu(yU|vp9_8r+8pQSAP>llEzRFJRwkJ#OmeMg&Kx+7|3_9FD1z^2s-If1Y}%c5*>1!=FQ`(I|erbXgF$ zS-!NvxW-Y0mMLcNZ~yeamyU+>UgAuj9o!t2)v{4tw~e0QYSqKpz-%PJxJ{dc=$7V?bu1J08}gU zApSZj?AysmR8HEaZg#>|(LIG#`$*JDYy@_<($>V2WH;V$Cn!<%MEnoG0ChkOmN8t? zRuR|Rbg(p&b}$IP|DQRVqFi9ZPYq)y7sQ45YthtOD_Jgt$3L8y$~*8Vs)+GsC2?}u z0hqk!L^fkwmA9OEETT}#foMd$FU`ZZ+TTX*8r|@PWBs2Zv#JHY6uR_bZ}eSn@#b4T ze$6MD56XqDbE_>sQdM?xLF~m}Q>L*}`OT6yWeihgB2wvgwvon3*VQherX{P0fw5l( zgc-(GX{=%mN<(croB=Q*n0hcIx2%|lwV5-htUjSXpS#IwzTIE1qrrOT$8&(TX1(qh zm?wQCX+u#j+gFTa2JtEfY76QKAk4C{xMqaM(-da}O+0<=Z zo8$3qpwm?tsV%KwT0I5?dbtzHil;50byRl?fMc2w`Rj{7#mh(#RANA3;)Wnys1-Ku zsT*%tH)ZIufr+1I6JWn(JXDviNoqv~lxtIlAZxJ4-1am`A|l_zU3tnb&zE_){jsnc zYik*UoIpwYgKye?aJ9Uvm)CSxx{7$w2G93mQxkNGMTz!V(AhPNo8jeq!nr;4#eJ^sb9zpm;!PmmW|Q;5}b&;zC{8k#t4>{s*<(A zwH*$w=%PkL&R_(Lo<(HSH^J~M7vGv2Zusea+sOs7or*ufStP_VjE9}4!!jc?AsR1> zJDnSaJGY#&l0y%Awb65T6X#Uy;TjkLZ0QDVq|7893uF188*uB*&zC zJAAF>P>WShK}24MxzYXP^Dnbdh3oay|Xj#^->mjlF>eaz9&j#u~M zw?HPN*dt*-$PAEMaRO; z4QAphNYVGCQ}INjIl|Xq9UP!aB@*x zia$WI!kWs*U5b;peGfTTgsg5D;VT}GE}z4jBnM%ytGeRQWSA7Xgz!B9X&iC+zbhRM zdjaD<(<#$|U#FM3XxUtjrv|88j*)=48Q+L zTu=SLSv04A@k{WPzZDqNEP~XPr|PNJpD?%ctWd(qMe!E?FvluepdR-Yi=ctRDO5|a zEnB)Zy^551!m_eO9x?l(JswAn%A2lvJA$eR9f}!%(tC3}i{t14PD-0n3leS=OXA@2 z%N(4VyqLEVp(<1o>3U3O=vC!3mZmR!y4s9aDO`0g{`U9N`4v;#MsLQ~j-=2*dq;`b zKWfQ^?YiLCWX%Qb!%MP-mqAP_{_xOhtC>~i+%TIX2LsXOPFjMF1(G9@#6+G9jgrO8 zz%nQNS@{UybED+-cb@dugKs;e2n;-ph8TPO1*x`eX8({(X3 zdL~tCyr=QbEldpT(1MNfucx7>a|S-_iPrNeG2>XU62=|m&0nw+wLjF@l;|;j;5|(( zD%N~fQmgoi5}aI=l)@j9TB~I(`vl)g#AE1lhi-^G);!S_IzW{~mBRG5U<{^#xNVgo zaHsI=Tz-e`UYsIZaV<~Bv{G{{=HS^!bC#z(1YU@}^%YJoh;?|Bz^!HO<}xuN!J%r0 z*~qaQ4Fj|I^1@GV=Y!%m3ddL_^vMy5gaR4Cu=2D!8oX*0N zA~NPs1VvCa0gW9Frz ziNJFLj;mDoL%#|6R3)MzBi^DfpVL~NB8vPnb$~|=+*&JZgA(+`G5b2z|B7L2yRvx5 z^Wl?RvUzj^rYM+9#wuMv!cIb4Vieqd|0B-+-t)ZloNsbzS|~X|v!4u8IXrAhv;(MI zQ837PYfs7JpM^2hT17!5eM;e0vk=Jz#sG9Qh~jJzIa4adnHEmlFsO6>8v?djy_V`` zv?o)xE+RUG;Q>V*1@j@=0sRp_R%Wmq9nD`FK)NCPWlLN!f)6f5>J%bF<3v$Ic!!jpz*ddphwh zabyQHbAGnMV_{w*K8hr_8{D$dRU4(XQFaR12IrFA;FJ7R%F@`OQa-UlgIGW=QD#I) zu$ALpF&We^JE3mP4UuRj@8gR|H-Nv&TS`OiJH;8sifF!Rzz9RUv7w`P^PI^@-P2~k z`U)o(B{%g4hSuwq3`)@6px2*4-{4p2xU0R&!X-4B<059;HpMVmR61bLU8V)57v9ym zv&6Y&bzH>I+$95PN4NcXY*U|U&xSE7#rv+z{R};?sAjXdmSHp-?T>?#*F589A_X#8xaiDx*tRR?KTBCB8}v;!euM7pZJ(s;`3?(;mvYEjx9HGe2#76>E+ z8U&dg8JZEBuxO34j*M^^U8A8qQz9=uu8|;&oz2>%?e!k+3rT)o6=B=74)ga%qfZ=OFFwQlEhj87*)z_ND>{v7I0AIpb+Kvi@Fe|Ud*9k zu)|D6am2K|mj};jWhBy8FE(l3too9o;pC!Nlt1LKnkAWNdmcL1!4*;?+VK!;XT}Bs zT(CtF%bpG>{jQm2v*!AND@-M2J)gDwO&&yyiQUAN*cNF3Wb3{zsVEw?c^6|sB z{ehb~!_WX4I6F@Flv^foEL$vHZX8JC|LQFN*CM0LIyN^NdyWyh=3K@63cde;} zXr_!gWgE%t+5gYnyX?l1W?8!PdC6$j3{L{d`FDJdwRP%yfi=q5Vd z7=DpNjrtvW+6RpU4$voYMvVq)R5NwHNIyw`YyW0`G0`GeGX6;eszUS&Eh1&x_P_Vq zYpuOI|6o$HU8H7AOA)6`p#QM5ZSatT7CS>54zyvW$2qF^Y;AEbVpxhI9DquXG(3K@_BlZqAT;=)ry z1RS!?cbl-g+ zDL0>uR+qv@5>bKS_U5@FpkD7w*myqwNEUI!IDQ6pDayU#{&_O5*25Kq;qW7sG*1cB*Ad zVCoflg^8-}=p!f~59^O8167&Iki8jw5=osh z5JeS99_LJX{*@;zzNrQfjwo#aeDy~H_!@74awt0=aK(_DHL@^1o@Arum6kRh1Rtca zLD6-M>HX+7fLZuT09;pov@jVszGMKwAGhR69H_%j+N#7xVBk56pWkA zdvHX50GuOkHcPN81t#Rizxt!8*EaIhdyU=d{z=6esk8Vqc*&H-o>g&`CU}(EJlc*m zn52u%(3vXeoav%yMRJ`6>vv;Eem;*>--wi;E(&TATP|b+F&XfN<&@r&^Cy)f(7R%> z!T&lh@@&kxH_8CXC^q~}kZSc#j~h%5K3zrB?StRwfkcWgsgn6#P-xYxP{) zmoU7c&AFu}=2UQ`_4{xYBHAP85{Ri>^3?xI4-~2$((RX+={Hrf?=_MB)l!Co>Rv?G z*BmTQ;Kmd@J8Dfo<{kt^vkc}Q3JQO7om!)ov(VGDDJ%HAoBG+bnHmz`kLY5_h+aEV z*4S^0pRNizkz0 zLZ?-N1;JK!_%umw)N^9zD7?LinHD?9-yw2%g((wQrJ7W)Z#iJDQmSTw{zS_(w)L5s!zfZ z$!Ihhe)~7KH54@+HLQ{qM5^(0&Wg!iz#yq$WOck1EuFFyEj54WFPS%osI6fUvt@V! zVppE-R~ZmUEtBvpaWb?4r12~FwW3n3L z9nS?F+p|TuT zx##SNynM#C4dhfN`_|wfa=h+Uo+f>75=XuHw*GN zq(r+Bn%vU@C{>#2>o43XV?Dct-{*6|3`K8I+YkL6-E6xI^mltwJ;IRfJbXb%ue_k@ z{6F}$tLdu?q1~!JdXYjk|A{i zF%9=}(~sq>UAZhLt4BbxlUu7HdAEtI4J;N;Jm!&*I`q&8k_Gw;srB>mO~OjfRA#Om zV{L|m3j!u_hDO=gLK2((f>>l@!w>Tdqlt~O3{5>c57#=)96@!O7OE$&pFiV%Go?^j z?ci$mF!MAgO`USZuJb8(Losnk8CdD-l4qt@qDh-#9+)aLGm*ml)Yu&0PJjR}0526g z#SZ&M{U3MO`@;t7*wNkXdBPB>%Qy+BBv~a;J($DA8?2u0b}!sGVa|YXDQwF8{l9SL z;R$@aCw`86{4Oz6n338UR44dMmlo=3N;=^4D44`-+0yWdDHI^0?j1zoIsZi$FvzW! z;vxlaW5d^2ve78ZSkj|2Fr8wK?@{_>GFt{Y=HZk9+aj2f9_nrZ$vNr~%0T+2?O*^O z9J__>fU%>IdaD>3Iq_t{8{hJE#y47Ju-#K7=+OhXUdvl6ff9_Qt%5HcPhEGg&bg9! zHcck3{L4M!Uj(7shiwf4)}V}Cn zncY$Ng%Ys)1SKvi6rv@P8lW)Z((t6SrzR|g&fe07KhUjS&ua~wQbcYxtgw0;O$>ZH zo7`4kjjht#4XFjMjnA?$^E4fP7EfTtLtki&UZ@FHDY^+H)D~mMDST|!m`o;RJ^Y1Q zQh`$}k2JmUSh#XY`o_9H9I(!wfxGhvl*$S7KAwF5D3S;VjG%!k%rtG^-es^@nGMk* zB2PTvj~fsOoY|Z)$EBH@JD@64m>x2Z9J1kzHNViKY{38-1Ocmic@}=;C2RDw1_>Qn zFrgBC5nSKcmrtJ{3>aVG2Vn!qi-O%|v+nB<;>u+S9)GYU!IM*ouIg%2tU*LmE*_ zjY=l8dLb2VRc;<9v82%lDt|AldSN1^g6C7qSF_t}m8RMChildwc|=RQuQqr_Bfopf zts3JmcfC}qC1GG+(J{8-9VGDc?mjpRdBdH;Zo}2L0I(-lvprv5JOVqQUZ>OqkAC1(3{cx@*-rj3RL}oJsa_-Vy8&!KD)jj zhIHW3Be;%Z-rdc{@}?o8P|BT8K`IhW*Xf#=%g%QBI9FkeM9>#`qG69gD9hH2+gLTF zmVZ(M_Q~IRc4;dhH5vi(Fj>UhgWjgqxttzPR(5w6ylI{`b+m;Z1sK*SJtK{{HB;LBckw7 zqx6E|4= zo|-Ke3;8MsoPN_VAbY%y!U^NxLrveqNx@L@!Td`$0&3pFY7 z@xly4!gLCiwbmtb?qJ@Ar{fI+f3Y{tf^e#-ty^l7(GTyT@S<4h4`rG69vSwhVirYt%4!Z@HJv-H%b!{{_xKR)bP7oW>Q8{pUy`h zT)0QbC4HGWbPlOjv9l<(C4hyGiSeBlvcJylj+?TEB`4zmExZkynkU2r``|1zrP?!ulx%*uh^rFJD#@ zZ#T}Gq;TC0XIirLooDC~GqdP2Bv1?or|+GzU7UXj9nEFjM-`@HNB>rLuuuR~F;ji; zhrSX-RVHr=CcalNue7&3`B$)%s4Vg0kA&KQg?e{hVRCv6PX1L=&_(S!i>1#Zom^Mm z4Wk)8Iq?@4EXEyjAZz-Y2Cja09FaBziyVQQ;36tBNJC9oti*Z4p>u@20e_s{j_HId z@17h!VLDuJzTe)whxFuOU{jA4!RWzrS`SgY>c`LD?mdH*{bg3&#m!2X6`m7XeQHnh z$_=_Z@xMa#3wLJ%vxC9;u9<7S`Y?&AU5pvIA;AY2A9%P$11lIYN5e##sfrE+egy~m z@G2Y?dzw2N_h5!T(2eq^uZByLGl|-bE@56@(&F(Xod${b?g4+Nc1AB(o;cUu+wp9O zYe?c-MmE+Rz;tsOPUKehU0#+f)>UmL-|AyKW`HAXUNCSHt|{;YV#<1QfwY?ej+LcRiKzFJ&k4C+X#%9j&!!)4#iC2Tm z@N2SANp$QDRux{y?8xh~2ZwAus2tv2kZ`I?DyJ(C?NXWJT4M`mqxphq*)}K>xg&`w z$?kdN8ISYW8(#vtF^@j47>=4R5eK}wNa(*Lg~maoEaAORiF;X~$eM7{s8fJujx-?Pr5$H?z?QU?iW&K@ z+F1dnZAXYR9@>uGSVUoG=hd)K;m7^EkG*g4V?;T?-*f^RTn-%cE)1?lx7FVKhJpaB z5Rh;du^(Gc)#`pLt1I5clM#y@96gFN#^sFMgJm}etF|2^@tQk;4W}uV?aMw4!=qdE zc=qamc8)RD8?65=DVT1A^_xX%LGGQW#hZ5Jl04P^<`j)qEe{w#)7fLE?n2vn0~0$j zDO6{d%(nKvu&H%KJCJA({~%msNC^Ai&B--$(DQrF!`(B7%S(W3`UI-OzMlW0cp~@E zlNYXCxg^oykFnio5x)jIdOX8KXuFY{>VjGa=J4)4Wss!8qa)!i-~3YA}(sm*+x$$RvrK z6R#*X)s!X!%%i}j89Q^TQR`YR`EvRp-Ooogz{dsS=U$e$4^uRF92m6gIl$hF_(g1W zovit7t&7`Um^(t^!Q?gwuH`SABR=>9TtVkCy`2@>8P=9BtQYXsu3VBN?h*TM=kN;m^jcs%T_s`Pv;-F{i9E0Z(JM+f?zm7 z=sBAGum8bO@%R4*+JDO=U_Qc&csvLSF;%#So~&o>$|Z>okH~&I&+ooFz9o$>?5Bx} zH62i6?eqxixBCTmnLKZX9E|6)8Oa8~5eouPBl~K>z}fP{R3UJ-%D_ml z!0`wtZsl2?Z%2vRDs`0Xj#v_9%bwQL$M@BW7_=}3^0sI^L_5(%HJEzEE<1^2NX;iF zL|3uEs@LuTQ$Nso;SY^$vP>_BF|f8Y(%rhX%+jJmI?fhm1Sic62avCWM}j)U)RIGP34eWW{ab|E7OVf+JqMGtca z_SMQgf^~6y=mcCSdra5~p)Qoo1urGZ(J<=6Vm-z&tY|{T(1L;wt0$FBd!t?om5rSO z{s6`%ZZmghiM#W{PB85Uvbb9+afKxsY30?$M;g+^5-;smGJyf15h0(dNm7MXmP82o z!jNdo7qqPXcs2vov?3zYo`7|ZLoMTi05cBpEK*$A1U*5XUMj9cOzOo$cyyKSbu_sg z#)B!<2d;+Xs^TfB?W^#pw@Xua_yaDR^;XWP+}aU=T+pMEJzdWSRI7+Bs!O6Q!Dm3& zaw1lNfOZ10@h18bj=o1jmV_AE7fDPL{u(FF%G4l~(^Zsu#c^3hU> zRDSOY!qW9EKMo??;jJ>ri0uyhLra^DJV4x-@d?7QV7jU!Neh?t907g$5}zwNl4Yj& z?*8i`=LFJdL-$@ktsX_0)5jLgVz}tW=}t}z?;c}UnXxihEtVS^p5!>~4X#|4TGtoF^ORzN7JR|q6d4OtH5zEG#V+eDhSjb)`gRAZjV0O` zEm+o>U~+S=TWN-6z45~tb{;nl;x1|=-AtAG;)YM34G>8Gmt(fuU{Wjwn18lqyW|=1v3;Dv*|h$ z{}f&yaz8UMy1t-%vES-}sRW~`Gj5FM7ZUB<_VrKrJBbzWGRsZkYu0n)@-Ch@<&Wa< zPN}k$Ex2PBe(n>ko&tx{ZKOc=ya|~`1!ZADIfOC)}qt6MLY&;#ze*UMy;O9S)YG8zD zA}rd!He*9Cc4K+djOm#BLbh>+8R3Qdv0&b5`zn1J?Xr|U9!-LqjjbGp=fY_|Il7$2 zczJFarNhoECO&uUxqfeoQs7cMiYRnY6kFFFm|2!JM7zMxa#4$M;y16s~pt6+Sl=fjt47EV7R*z0TwtS{BaP-9ezF*J4DDEWl0w+ z9JT|JDn`%=K~qqo58;bePN`3d-Sy_iJ%k}YRDy8jvK$WnXray4X3kj8`6Z>KEvf+n z>h2_uWKM+hIpyVwGjxR}zQ}+nhBD}jBu03P@JPR9du(;wz0g(2T7N`tpkhStYv89mgkPe4}lme1{ zeITRBz{jY7ORb5j)x?}t204t3p%czlT4`ya+?<%WF`nUTid4+G@CODQ4o75{V023< zn+d~1(axvvM?+d6HBDEy+8!AfD2blzQAWmXl-lz?bWT#; zBO2RjcthTLRpM^l8F-|mQbny#ZMPlywjf_qL3U{ z&%Qb|rD$#1^yEswaa0Od0!tbmvD!|aJmC=Wz?_Z4+;;Bb>v$Ss`p8m%StOvt+L!FP z06t&Yp2YDLDWLmsE=SyoONu zaDIo?GSljAd<}b*!oYrm;jsCfWFGtNSh^EiY9)%8lbe83*r~B=r6}WG;>X0RP)xq6 zdZi{&S8F>HH4UadAOr}r6 ze*e?I);H?)r|u1_ag(2))&PIns0@8g-MMl}o??&gyv@zL1ynn8imhw)3uwet!L1rs z$mGzb?Opj{WS&Ey%Y3Cn0$b9P<7b4gjoH3#N+uC=C`l=%?Xqb~AnU z^r@5pPo{LxMMyK?!ty}KMK*7>sU@_42~KBDxfF-Zo{E%&eRKW{XvIqgrxho zE&PxY9bc&T4kVg0?sRVh6%7JM?lW`3*g$3z12TFrlZgjl{UlgI*_l!+umFl5riy`u_$+ZM04Oj__%h7G1Bnl=jVezG0b0E71@;HY zz26V7Nf?C3YrHTgV&7arN-vT%+TN50T)8Y!;13&YvGV8cQN^3_MFdR-4!N9gS=`25 zl%Ki`ObkI#eUhc0XUfQR4a#ot8CtJIiK}Xq2h*Ui*yNAf{~yZ-rO_BP{Go)cMlEM( zrLhvFLG?A*5LqE5eeY`5S8PJl6Ih&X*icQrd}#=x{RVXe8zdPxWpf4FNv~Lt&Z0`b zLdCH|zLpp`H$!_z}3YixitEwTmA;UUTlExSZwl{py`KsJ`P?ZTrEL|pSBP^0L%-xp#lh#cH$4GYE=%0_6Ueukd1T(bQHXSnb7$|% zZ*yo9*zs$4eH{1Yd}UyW`K?IbrtmpoaSUs-bA}g+7>EomC>ULE-oUZdXo54y)Q%c# z=wo>(7CUA-HD7)_T)8YE;ZOSBVlngGsU!iwDj*+nOG~C!3V`Qh2>>+il15-R`hZQp zq-i9QF4p_b<~du3K)jwXBa!DGw_X9E8dWPAfX&Q5gX{u^Ha9;;Wg5Qni~3pdGx2V= z5GoQY9dEwAc4ZeUj|qNkZGggh%pR@0F6Py^g6Vg%wHhhvR3^~rUec2=m;$IePYuE+ z8yJ)`bU@e4xj#^XQh~mZVeih9%Ij}`n}lD*%w(a*{^Z-g52v$l|K0}_lQ9pzJlO2B zr1UJFK;~R&bunjUqo#OX`1*{)>w+;;^CP+^W@;s9eJy5urk6bKW>0do)Tpx;;?J+0 zmVYp4wSDK9L4&S47gyVNK8deVnGvMB%C_Egbh=yqC6|?>2pkO*r1HXqN2lnL$SOL0 z4w=lNFk;@qC573WLg__h#NTb{lSo72t~=^P*r(HT;}e#8jgMXO_j3zhO|T zleda@#kqrqSIo-^w^R+UN!b$n=gNh{S3q6ZzS|$3=pPPqGrP#0xfojCxz-D+Us1?{ z^p4lm_eUY?@gQo*5rTwthvq(`pj-mB9D<1Pkt7rhD^DUM!YUOZJtl+C!ph$eh*gj| zmCzcfpUqI5d?;(#sd>OIJSw0?*_;x563^o7vDck21;Uqd#-K-RfEYHwegFzns2=84 zd7|;zl^?$zwjl60-`UKoam2B6}P$jdqN$i%nQvI4WXHdPoXvN7t zc2R1g45GVFXHBb=)X9mPhYaQis+)J_e1gI0C4LaR((sv9iUeYAL%*tea&$4O=S!kX zzI4xl17qN_p<0G*B6h98ofUAZKQ%;U&u ztChoIf_sd>RoTguEaT+N!YxEy(?}z1w~!udbU{6DfNvN~@crHp=RyXG+-wWU?2)#2 z%dfwr&;Lv1IcuM2+-7KL)xh&&JHYok@mw8Mp2K=ASnbsAvq7aWv&o)(fAb&!>m1d< zyXQStquIu~i^ak^ofnBaxvJQ`FV25#+eV0~LOyetyo<(cMB}hI{`POPuhTv=pp)PW zS_EbTW8INp0x`X^{q5f&3tWe+U!DXKzKb}eGlw|}27SgP!P`?t}^6etwTvms+k z!GAY5#SRV#tzPDdJmqDODp~mW8xpl^&7AN_@cqTUwiKyl^*gjT_hct}d8=4nV#_cx zY17Koz|NVZu6>S(f$ln|Gtzpb<68i|?(KX!WTPKCJD&XHSYc9deFc7FjS=zI=Y+X{ zENk)SBu<(uJOsmu``+?1p|&fR1?1cK>F3>7&-JAL3(ueVI41;;S!(8dgICxn{ORX^ zx|qN;4=Te+_~D;_mt)Ra?a|YAy`EExtW*+HHgi>wOa{24%6x9(0&2mnOr!627k^XX zQnYGTUn+SaJ9rx zMNrx_Og>~o>ZQSG&f15C%$I!Ez%$O2z#R8M+X{BOwJkrsD?IF5CE&W-@lo>!&$jVI zy?gqTZEfF1lbJ@!XIQZS_uI*E5>10NOsw^I)$fny8Vt!6doZHrpqO-i+u2%CSn3Ll z8w0*F4CDaG#El`o!_S-!l0`pY2U%Jym;0B7CUO{o_4`>P=A#L$MNEnHz7Cu(67l!+*t4OK7M5B)h{#CQ(eVNoO- zL~RJ1MpwzKXt^)YPC>U@h=eEHMCadC<*4APR&U(P;SX-=I!d3w|}n_+b`O5{Go#g|-lpJ?+XT?vRS^x&BxcNaX?%wyF{vy{$qqzIeHY*>;axh2E%o5dL=146skvPTv;bA! ze~$WBg(INjs|m(*xGYVuKlHTS?&LJVy>O(HiOgZ`@Zs($Ty2x^8^oT7442>H3&!79 z%)F5u5u_9l`_xb91a@{NEMCNOBP%3KsSvWa@N~|kBV#+zGaqptSt&f^XFINX@=JxL zP`G=pZ~0+ula#hfz}j}H!y}&C&QlVm=q#`*XzUc8{{(6!2f;BmAEHtUhI25=vMf$Y z7V*0CYhmRNTn5CPBED1wLwizufXTfdx{6qvZHw)g>})hP{P>mGmCI6l%9^dNiju_&64RnONH#G=U7girvdC_fAZzEgofft>Adqc4PUdtP*|{+> zLp=@linT(G`aN*`he6#ux*fMWdBpnNI0n$zwk50$AI|)83x4XgY@;Jz% z9$l1pP;S6I*4eBVpv$mD7dw0TjWK3=v|%U}A%+p{^i64j9Z1IUU4Byo1Wvi*Ci{aK!=E0~`+EX6bsWhgwL*x$bG7rCqrs6@y2+ z^uSiYo)1kCX+<0ir~U zEe!_dsMcUx@4R#%azL#C>}A=tya!{h7*{5`}ku& z1Ckg=A7T#-qywJvt!+_2rZ;Hq=FOuEJ$D!w+kY&UkfB zoVu}#$9x)c=)gq9iBY3J#f=LWh1Jdc&UhiKj4iRicb|y*-bF?TZw&=d!H{VCn#eWV zWtqs;{lTIX>^U9H_A!Er?P$O3vO5oZ#fg zb5R0>h+?=XAe1uKUReudKWf@b9YPxn`N^cl2L`b}pM2nhJyy3D44;l4+hyvVP6>Pq zTtn;+I_=bQVYy^a^D+=BALOsbB0IVv5EG}JIAKq(_;6Dh=43+UW%NmRoeGuZWuUDQ zNr!*@rJ@gLq{s1$f`kzGpt`s}G;!Z>$!JIGgWpnk6#~}TaQ8vj?ai%wNJJk7PxWZC z4CfFbZ>Y1e?6m-9n;XlE>*vgava?TQ!;~XAY@~l zP6e~h8lZ+j@?yp5KH#hr=b0OQ0E&tX0brG#3Pa8b6%Ih+1MLnv2nMPoy8K6dNhtBs zp1+P*K1Ky1%cO!i08)Y{raMLYmEl3;4$p<8WIl6TS~i!5k*Sylo*4hthgQGg}z#Pj23i)=M&`YIN>cZqM>OKp$j;GGOuw@C|DPAWL4J>7u5V34%` zarJ>el(Msx*B9WGU}&WBIz`wZlEi@<_hA(EXTl^;LvtPKhY988m|~&Yo#;!sDgAwT zrPK|QVE%@7*GLO=22dt|DXkY^VFrdDj+AvC!L~W)?SeOJ(-*0xwOJNYO~W53xw%o# zu|iz)JX{RNAmJ@3G^x3JB5p~zoCHeL*VD{%efqm9rS1$X=7<{bWqdj37;su8w}1)4 z<_+WJvY-KX8C07vq2EtnKPhqv&xr#D2@0%V}1 z=&ww(GEt@RgBp+>odkeN8i!L+D!8P4J*2K+IPn`Xl~x=yK>8(}2AY{CSqZlX7F9>r z@rN0B#PJ+IMSkTVjtlLZ>tHE%k8<>3&Wff#Nqw#=EIr0~j9J;n4tN|z!|9D?%j3RF z32uy#e7?(C4I5TvZX;)eo$kBbL3bZDZBqM~+)EyM-Amn9>IdEBVrNCi8zY}9m!vFg z`ojvh+M9Png*aPG0EZWG;rw;XE{VQD9l{a|vJ=CEgU&?#59$aQPTcqtsWJ$Ruc5iB z6F907T{ppWGG2HfTM_8WKBWZ$upm#LiHxTy1XnIg$oK;nxALq<6PHJM+>dl1z@!sl zds{BIZcO!N#BoqPCcgkF_VHM!49x~hT9;-=SJ2}U;E?Fim{D8kKo#wjn2%AE>IEcj zt$@;@z;R;XA|Zg>1Qk@g&u@RUQ-tt4<%JouzbKT2HoduO+m%ZYQl09CM_27yBaf*0 zs+ZQQW;i(u=wL9jjprXm{0N(uLAC7&>V=nTM~y)BhY}R15ULnhAq8YbELb(GUX{4g zUg7F%B^qm254Bxb?^C{-!K7bMQV+D4%#FS6_zJ-`+YeI zhfJ0F!axf8ysFOGZSitdBL0R z32)%0r+CAoI*%(&_?vIxSeSEyfGd5N0XJZrC=0h0!-*c_(Gjvhb_Kd_C3DZa7sKdd zfYCjxG6wjvstkOPJH7c2JF-2_hgzFCc{%wTrZ%sN0}?YTqD&WQ3?u;7^6tsW_eOh; zYFND&j!cqWFaSu#F$r5xfyRa$++9%WQ9x%{$UtZl@8o-^v47~;=E@~W!v4lZwRT=f z{`QxB{64lv3bGT!EJ}7-J=C`(RZ(8qv7Kh z)q1pZv~fm1(kY-r-vZ1+hk9TE3NXZCUA-^_`Mpd1si-Ol5Y(D}Ogf37W*JO66+`|e zS~XS^-VIYO;)Xc{p;A2yk?3jUup56KPh}0R+oy>H(@-JS0GF;*d~-8(df`m^Nyb_H zp#V7#49-@M@E3kjK(dF{LXft04v#Ci^DA0NPQLJ0nNyjuNBNbo{R*D zoKhGGC~|84yqs&DMh?Qg2pzNk1&tt9`q8YikF9Wo#eaqYqZo1-=Up(-GU80!Kz{yA zsxtX6hEzn|>swhn%dz$7RxQS%23cg9g7M&KZ!GaphP|=uASE9BO~7iMJaum`yg7b@ z5{o~1yoFrb?XGrHpTzH7s+MMnE;9CnIXgFsOmyr7W1C&NN!SShhRzm6138A}j}$wn zMw)V2C#_th*P?UPaa6E9t$A7-OW2fQZLA9$j|gWc7sj$m*xi2>v`pVxE|Mu`M~+cp z2uHDjDm+t9%Vh1!CFz?ynnKq)dFa*2+|tWV97BzuSmE)vnihW=Gr+69DdK=f|J>IE z!wOfLBYtwD3KhVv3lhZ}@uybJ5>~}dp}ME_jYLT));HQILD*=9I4%ADCS&0)v+4nr51rs9qH#JC6{;3R3PW{T<`t%QH&P+2;=~v|P_r~rHI!8>!Ge*~zQ=Zg zhnbAl{Nb#14%E9_*F|0F^mcn+FqYvjQ@5!{XKEE_qTqJbMV2X%xqV<1Um15f1`W@$ z0T*^_7=C;qFx+{d{`R-XjtK+Ej=6W-U#lHh7f|q?dD?-<(_V@li1xJmLtX0|jhtXv za=C5g&>6m&Nvw#R5A$JN+v5&`-EbstfZI%JZ(_LX>-gZxWvNE}fv@#i-g=3wcJS^Lljwp&J5kgUFXaO( zWk@EzXk5O!0#va><+5Ruz4y=~>)lfx(Uy9EJ| zhBq1yS1w5=@n{QOZ`SiV7R2Pc;Y3T=^x$5mYUEWBa%zo-EV-KawhAkmJ9s`nm6(2z z1IJ2HZ1VbQW|bf~b7a}dY2z};cB~4g@R+9Ity|&BWeF00;41qKa#V)G5u6JBiiTydocsL;W_>DblB(B9kxVnVD0|045Y}UXXUa0BK6o#`f(KEp zaSGh0U-==z0aq?d?D#`i8?{;vA>}MFu`%(%4d#Z{8r_X`7-6tGvOUh$POYwl{J}b9 zy&yrdmXs!Ix+4y-gem-h70afd4K~E;X0pNN(J5sfy`vlTTxIMhbxcmuvhN&PRivE6 z#x~duFGll;ZN4%-%gzMXv?_8JveTme&BT&&h@tjwm2b=@tKbT5+90y^@M|UTFA+tt zooE!pDtz5gNRt3}4SK4tfGz0P_=0L=R~0>6V6pPjivng7@8L2CY*h+-G?s4UmcskE zCm6>`y!XXMtaLau#<5lBoJh?l0h@0ukC_hS`nNZ*}4#YTq21gQ5keuaJ*tz58E$& z1yQ{W1|1ee{*c`U#nRot;92)IYwyoki&gDjjzyC1qnA=AwRe3H4hAXmj&Q40pr$(W7HIoiL1&#xS597H;*lxl^4pSf% zjG)Fn5Z)iEn3|2!R80P$Wg;kNHn7U9Ln3v9wd4Q?d$^XL5#juCH?Fn|j96NUcWh$g zrmNpt%mkoi0`Q^G0KYJ#cghRs~fjUQ#bm9 zP#f)yyz!1TBnR`zg~Ou%z{c&(_=+7RIBe}F zi7EvWXjNM>b>6eDrdVER)qK@|T-o6rk7}@@yx`Fny0O{L$qT39PK-IzGw?7}vlCiX zXUAb8ic}tNsQRMaSQ!t_N3+N^s}e@J8qxK*netEtAl8n_oS&F506v-J8(*ID1+z}I zrS>kveU*aBC-#TY>W_<#j9YbIMIcu$32f%k6S}#P2b(2>;pKj;V`{%KmVeUVB+?s; zqQnoRE52te&@TobJI-9RJ zP9~peZ65V(f>`nObg)-2dbaK%^8TT&x3yIUUooq0{h^!9jdqS=uNU?i;Y}-Ys9qVv z{Ep!yQ+=F$w54|#X%F5xd+Dv0!leCcyN`HKJjm>X!vcjP2SzV5lFu^<23u z2aZ2jk|mKj`xDba?s_pIvTIm9jW6cRPMkxSrYBW4=0#&$wwQD=yWw>lO}_n&vm(>V z07x1)*+yt)5(IQjt*Vm-Rtm!wGZ+va)}x=X7mH))qsV5z6x8>&z3Cyia#>==AMur# z$svyHo$kJTR@wV)9(^KAXoYEtL(FN>@$R#GWU*LLSGYvxe%exd!qqyWtRya z9~Se6kv8j%oWh}ddRFcAo?#HPgfvdWkIsg+au)W7;|PhA6HtWL(Q1jhG9vI68Icxtf(eV7_t>k6Q2c_SDyx7Qz*kOq z2aKhV)9~g-d(#&n#+6GFKpu^&Sv{OnexytvQ4!aay2KQ!@Aal;?F*Ie&CMhZ*ow8} zGT(`*ssPc{CBZ`IcU5yFhm3~n2|bZUmpAE!i;he&ruIa_;L0VbJ3RVf6D&CsnMG5( z%12$dgm9j=4j)@`#xkeTZUBYdaVE|eJ!{quUucLiE!>PIo!yx`zFwpte6g0cYqgcD;ORPKW_YlN8|C` z_+75?INf2qb9ZRK>d+^MPLL=^ec$}Iv^L#UZay@8RIT*zkz|^;}D}->$`~?JbbgM4p zM5Nz@2Gr{hLzq4s`M9TtvcJVYJzFoNnAh~D%pFtZpoJP5&C`b16ODjQ^AxkzJRvYv z95O``=Kl(zTRd@|Jdo|mF4aFG_Tgg3H;glo1gDN!_qm;JLE2Yo%IvHHQCl z!S@L;3_^@AmW`WCw4BeULx{e_iqjk%@T{}>-3D+@-%k~)!@&tMHF`4_D>t)UYM&HU zLvU2+V;rnQj+tr<3}lS26i#^KVJ zm0Ngs@%cNeyg)XtsT-2AVofFI1~m$v_mR*KJ4^uIov`4nm}Sfsp3i=Uwsz%`l$sug zFs+T8{9~5j(I-WSAoxYNP%ccxGF+l!=>msLevwwh1Z_;~Io1|^hVq5EJtq*E^&QD9 zpRejf65aiam7yvz{!-bSTK_1S`}67x0_AL~84f2%^^Cy{rmC*~498V703aabnRr><;ec`9<)@&^ z{FmS6XtlZ%D@~2a?^FY3H z+@a1r0Eq{>L;kEdnwy;*DsaleSz+oO?w+2#rB+`p>ePL}#W>QA778V5amW-cjCg}HMihq#f*AL`-S@QFzD4GWpro^}%$ zz_y9249qypRrDEJaX1xIi&CIq7g?|QI@Y;zNxB4&4#&-{T8_T;=vR|10b_~lQ{9~q zA}m>xx-o$)RvB7O+V}^+1bu-NMc6pvWwR-T(>_w+EcTvm#GUv;7Y=PjKK;?tQ_Sd5jo|h3j zIam|{R2k_Yiyfb89+Q?b*$qoqX{65`#my*H zHCLA*|DjN-7dmm=cnmq+r70fMP9YdOrjOevX+;i`SB^$DeH6se>uq0k3|B5oD0u`r zb5~TW=G7KS7{w<^Ijgr=a|cm%Rf1rp?=+k_7ON>#oP5Q2#aV)&mQ6DZwUA+aYB7K? zmSuL5Nm^|~)nX7jcouNE!#mmeVB&@4_nFWHY{(^t5l=;fXwzGnvss!l#~&Kn+Q_r< zI6;Rnxvh5RwsIQM=vuk6TH;X3Qqtb-w!So0U-q@k;F9cOjg+{9lsUy_CUs9jCXS^N z2$_tgCWLU zTF!uLlDyz{Z^ul<`dSG`x2&74Lc#e9%EA3h>frwSX1F>QKsEVY15;dSqJ7qt_lCSY zcij!( zZw0Rr224{KqT1A6t5#jn;8aUiv=&SaEkC+@Zh!0Xb@$2b_XohX(D~jyAlt3)RD0~u z1z**AhD)~k7|n+cWAQQtS#7ksg4DXxBHJ-%qJ~o`ZB1xwv@$mtXrP5O?>|tLY#I8X ziKF8SH)(Vp-x)HA9DguutKH0j?hj6%pH<%;zkXIZm|&i+(i4d+jm!+9a4)m%lx3FE zOQKC;LS}RO)Sz8ma&x|JTq_NnWk&Mt`vK&MUB8n9|&HIQH9%krU_lKdjH}bZWQd_9UNNG9U zHA>EsU=hQk3C81a#pv8BkmT3hEmNUMO<^#sT3>pZ5Qn;JtocD%zhh%`@uHqjf-5du zIYP}e?i&U+cIl(bU1Wv!b{u{J3k&1PWfbuV`fXnC5p6-GJ0fk2saW>R=e;3EraF}5&;6*4l=Q}zUR4WCE9&n(1 zm57^Vk(Sl{ft&5dW*&lZx?@a_6x4M2e$`|{1;d187$ONJ6Oqh)-)F{Z{!Bdnwx`KC zGuh*Rk|<5g!l~F9-`co`s_ciJdahiOW5A+a|qI0hK9NYrei;*a$bOG@N?KmLC;7VD6)Qei;17qvvmjebINH_T6!Gy~2XpW;PHWxGMyxuWkjwkbkFB3 z1d%j-*I!geVCE-}zk)Z=lj|DCO$l7rk~HBDdF-^BIdZ`6=w=wa-Fap-h$>1_N`h;X z0ZAaV%n2!+=yH8NoOmA574(J@1RCxdq74H zl-2z~d7aHXUH9oJJOpH45$=X~kf?bJ**DWqOLf=Re|K72Vuw{tBM0CwCthqcwVJ3q ziY~|3f&fkYhiS=O@=Q$(lY=?+99s7pN*sn(yovZa3%(b~V8OfU&2gw*9>*bnKwzz& zlj&D6AEILx&u(ubg;iSJS=_v1{~g||n&1dC{=ghs6Ge30!Dx1ysSk`Q&2#TEJ4Hd& zL_qis6odxb_n8hn(39;Cq3dkrlxudF3W7~5vhW#5?6e=a{bH+CT6%NTsXJ%F-S}F4 z+kW`j;XABt7kL$lb~4~2PxNEa1->8lIZE35nBtuJTL16P<|=m*Jr*Z*Poh3)`VMi0 zc?eUO_I*n4A42%Ja!HPAkAAte&32BhW$IhWglD#lhF(bZCb>i}H2(Wt2oCZ*(>CD3 z@R7|xh_FX&&$G(vNuzpc4E3PqNCwvOSZ#(;z^?yK=AB0-@MB-r2avc;{hqc+(%8R@>anDV5Xbj}(_E`=-L0P0m8c#TfsxQ*=el(I0Qpfpvgm8~D zAwrkrit1zAlVC1frmq(o>WH3++#pdG7_$$`c|s>nza)@G9)tE}#0cQ4Nq6RCOdJ6! z;0iqL;1~GF{xND_>}juWY&4tInn!EkdaecAs!V9nPC12Hg1hC`XFFOc9etqGzM~+c z0{g-|jXTsGL(O~f1Zjkph0?mu2{=@!=>dwh~cfWLHKZ`E?rJ~6 zk+aItvV!N=7=|vKb0ko%QbmnYfc$sLaV>3^m#?-!SC5^ABj!0aZ`bgBC;ZHtMTDlh zZGLEU?#hpx>eGPmnCV-aIc)!*Ua`5#AoipqTfOruEto*?BtTe)^MNkSG8S+6w4_!$^q!VTc91`U31P+tl6i=oMvRB zE_tlmcOWF~gzcwVm%2AXR19tkN-AlN^f`;a;6L94tN@rk4g391|61Ru*PpsKEUuED zpVk^}9;w=f7o!??W%sZ?8WA?l$^ND(jokH?yYc)2@+%eqnsC}ak4r`*PA4N4+ztv~ zd;m;96JU&rE6sYJ_B)Uo*!PSgjxgb$Cf+n9r;0e6n)44neLpPtRM+ZW=AMd;`eV;M z)7bDg8fnxzd5l?{k`c*F;_0;d%RIzBgbA58)ET(Ca5$px#w59rG>BLR%3Hvc%=wz<#mM_V2Lg|Ak;)?{FS60&2)N|J>n1ruj*70Me ztH^qEGuGDRuh-aaD*S?2Td~wMAG@Q+Drq7-} z<$ul9lj)8Ar-qX!jI*BNB7VdF_NTRtCpF5^r(2!PYNLt^Nj1GW9_?@&t*yM7=yA7q zwujh%d2bvq>XdVwN4n37kJT+KddipOL)RH0=T~|7Z-%$Q_^R@fO5yWBYrz7vcFblz zu@akGi!mDBtDuEfm7`EgvZ*F0X6D${#KTEJ37ziWeav!ysK4XNB`N%C{!Aho@ayk> ztQRG|6_3^DW45L%RH#AI=hYtP#%|@cE%MnJ#q&YMHG)~nB$mJM_ekx;$Z$S^N8M%Z z6X5J5xWp};nE>ViA+NBfxenx%H`JsyKx|p*qU;@Tgq$!1WVIacUXWt=bQrL`hF{?G zBFZcU(UupDhP!fEg2*4FND$?;k8A1O<(*?FJC%K8w&BcxMtGX5t;@@JT%EeM25~0S zS@wZw)AA&`mX%4G>Ae{f{8o;Qx%ec62m(g;<|7J)1hl$%8{a$4ji z8xJ4HpR1T%5k70AU=hPeK8_}qsW{jPGdGVBVo4K7vl?+^kzeOuQr-x)tc{h3w|Oj| z8>Uh;$)MZk7D^;6Dx2=-V1O<)7-o$F3PxAmi*|%mRj-$2_)+r*Xf_*p);5da!72hv z02pD46y={Pw-Rn)sx>?cC*yRH{2~U^-_8V-ynW>jtWHLYL*(n~64OD1#lfcuC)_nn z4o%Q|Whez5AxIa^&)O0_nO|Q6nvCnY#KV&j9uar(lx*?gf~Uyoqaa%;&7D~iW%HQTx@gY!t}nuY>>M+JrPZnf zBxy#UkP{P^T5v09uSITi+7hEyXavVQpf~&=xJC}u#>yy|jqc+U{6O84KM=OLRnNJd zi|2bCL3NeP;x6|u{Xk*RtiI_*U8+N2^Rd;Xi|*MU zqT1Tn$hrTrhp$fdaK#f?F^E2Vm}4;WvNGEnPid4_STL!q@NDnp-kGr0MdZo5(STi^Oj4Uv$yu6&>iXe#x)*S0Hm4-*d zI9j|H)g{=7txE1r-WcF3U;d4~uNqnmH1SJ3#J5 z(9DK7Z?n(Al`cSna<*AxgDt*#HpI>}t7wJ%p>h->BlJ1(4DN3D2?uyCGPnYPJcYh1 zuifr^5M4q%qh1+be*4=f`i&fN9Ea0BfOhwqAuqqA}r)4FAOo zOXqDWQwj!2(-&&dY?cMJsQUvr+pRn=)#u}Z1Ro%n>P`)uNZ(k5N@9qzCi1aRVMj!$ zX+%0JNxWE>{xXmIEMR5KYdz3fuZx6hSjCZ5aYb4pgpyw?7)dQ( zlaFSrER&DAKd7|5mFKIHI`O)RhVcwpOQxy@+Bhc;cTzx@&Xu^zm_t?3l4z8kS(z;J zYBr8o*4LiByvnsbwO)5>F*Fsdot-9%BmAiPz-hgcW54t zDm}%)kk_*47-@>Cv3xK5$V&uh`sUjaRHExhmXM!D&+}(W!qXk4)yRHPav44+p=_`z%g0a)|RkgJ6ClF@qX2_3;-ZuQ9x1D;< zXm98ERc{X}Jp=lvkEQGNg7=|gXNHcMX})Rp_P*F6idhVGY`}II$gU6+UyV`d$+7%p zx0k7O-$%)wRn;u`REa6(bTQ}2Jjv-}MB^b;A}5UC&{5n_vm}FK5|4v#-INNdV)5p? z)%L|M;mT!+CV$9qClBQVn+ajNOQ-iTL^PozCmap3O*6(OpJR=KK<8{Z#Vp_#G#lOx z)E!k$^~nnrxrd(#gd<+<&BqtFa<{_z{HkCs+4Kdyu%6W2{YMvTdUQ0dHCs8{c$B;Y zS&$}DsW-qVFg&C2%*}UB;xQ%dNJHW9W;U$svCG0aAZIei3mES3{W!w{bba(VG%gGT zV`ZHYTYg!m6%kWKFxu9ZFRlnzE=vaRM;Aq2l4Fzj3h#Gp;5Fxq#zI8W>hgPrp({UD zDQB5DkP>1s9hTDy2g!M0^iP;$oD&=%Uo`YPFfByT=lin1E65{ait)#UvWFSpHT)r@ z^;#!K2etEhG9Symk~u!l)dVFJK!q6oz*}TreVjH4OEc*+sqxc|+Pg1bVqO{cO~8`5 zYb-Gx#+9Q;EO71;$obJ{Ar&t|7$c5!vhtN4ZH3bRVL?Qq-MGh1`yoVvE0^Vg_J@Mj zn|YQwd)@gghUE2Z#aO&6?*=&+YQWzEF=tn55opf>fM<>{qRIMxd87j&0a z|Iu)A*z1c|!XaGEN4>zmPcMok{dVgf{u)0PBdxNaRtBM6;>@zB0U5%55X-XEA)&-EgI)$a$_ zm=UuvB;78{-`XxC1lEW^nPK98> z;vY;-mJf5n9|i2E(v1Oao-wxI5=U$xE&8x-`m)VvZzi@Gk6xvF^dN4uYB>+<-YKHP zJ*;pj&mr!C777@uNErXgUSw}O>UPm9&L)wYPMxrm8i`!M;%;ECP4w%?qKrboQyR;v1(NK5Z0h%&n7gzA}?@R%Kn=*Usxq*yd$>b}l; zW;bOC8GnFbqut4A@GMVIw!!NL>5V6o#K8KE&8j#eFia6hgZqI1!Z$YqD6lp{-(!De ze4~w4@lD_?oHCJf1ln_1c}~YrXglF1b|v?*w*7%p&#wH)x4*WvxzVU@di2z7bXqwH zB?e#^>&ce!S&6hn^UJxicR)%3TozeoZnEp6K zmj9>VQstA^NzMI||E-NXo!sF3CJ|EkC{_cIA>h z&Hl*R%~m5P$)DkO_cl6%g2LWNEue%w5E&#W-x|4KEy!=ZYj`R5hzlA*qwb>@AFDS= z%D}^Tu0Rmg1d0;>6ol&@b{XCa22j`}z9m{bns|~Fe7f(``2SdS(0*+6KTRwC!2f2u zk;6gR`)y7SDL>AcP=s)DY2L000Fw5u?+6BK?ZiuP2hD7yzs)cV0s{$jER;3+`RpC@CTPhl_zsHGF&h5CQDUB{@O- z4Go%`d5erST5=*t1;O@2#l=`QX_SBFMRe=Nfx983<}@$35d!syenEjc)eDv*aX8VZS-bFkImwE{a3WC2;s|mOt9~N^j>wX9WcIA>piNCQy zb1TnrpBkgvdv-_la98_~iGc6(#A+5fA?)N*m7q#@RNMqBO-= zQYzR;V5;!P!{i?-HMnv~V#nWXqlNRv-E`OEXb?;wow=6NOoycRS)M$1Or0E2dfnbL zTaD;o!57cC2I^XV<7cZ**|?{YL?_eSjw^kWs3D9dBf!$rC!9w#pOmex$59MgH-&T^ zKm%V;NY}_kU{Z%L^$Vt?maiJ7)+$R4h9S)JCo#@L}amEd1@#{ zFGM=Z0hEm$5B5eJ5uWrqcIA>J5RZ1it;S{!2C_X0`d0|CzAXJNj>hq4c0l1$x7z4t zP98Xlsc_Gl6mZIe`OQr8P{?nvR(bbR$bhu+HpBvD78V69Yg_k#IUWX!s`&#C+qw2a zr#Nd**?g-?F_=Lmc?lbdDFAN;>AI!NJ$Qlj@@~Md0SLsL+0KuJRlOwybg zk2~$rGS4C2Dq64uPn*{LJiy`8b>#x#shnIPEhNwVBs@uwV*$MP;B`f)FICF&@8CtALGq@GM6#vifXrO2`8c6@0Ix3HXt%fZuITsKJ0_ErIN<+4=3{!qFO>&WlU5Zm3VRRrL3 zs(F{gKF$9u`>#xEDHMJm$Jb(1mp0(*BH41D^2!~PCB`EA@CD>jFn@2|1G{~ove+Lw z)@kLbX?oo^&*0F-a4}cNc}TOzZiS5_+^pOo2dNtM_KSoyVdeQiJ{2?i)=Zi|ilJ1% z<=l>@p96|$Soi}$u}5>flJp=)qnm(_d9WuRWfR%u>yR+)bJ#t6yU@lwaCc8SaJ#a5 zs2`2qxhV1c0b`x5ywSdM(c1eG&44V}BLkey53<-zRx9R|a@|q*h5L%l33|IZ98SVu zcB}C{?>v8^W)|}-_jdTf-aW>ui)r6LZvwXu4R6Okpdj4d@U!P0Rv1?rG zRD|8pbv&I_U*X^yUmEH6IJiw0pe~g@$)e-x>-hxJMkzFI@JNehUpW9~pX13@Hk)j- zKVl$)mYmcUH1Nvd6Ig=ROo3;`5`ClThY3zRG#{J&o2Ckn_OJC?PIYKYu9(M4$c;mQ zt-O-`>$(kqLqO*jCfr<@uc7{e6mr*SP8MlNZ=h;GC4qXn_b`5J(n4Wq@@_SIbQb5) zP`@BB9W2Mbq{gCP#I*0x+WpYbrO|%ug++^q@dwq_+c^=#UFu&M1Wu!VMa;kj{S3jM z5eZ8KSac~m@brfDE`rBt^P>Tf`3Y`BF-(waeqsciXBfU3AuB<&79j7PF`sKaA1qor zek^IoFP+D>q*+9ZKg6+-M*^_L&Q+WEi$6v~$wNj1tJyqLLAOVEE!6w8i?o+7;k3*nP_i9 z;d>enC~f7!$A1`PM{48`;b_Q$nnJQAZuyq4!5mjEOY!27l$f_FEh$O}7r9I1n+6*V z9dE?}yQaDVt!Hz#3c+ig-&o>u&iGY%BEeMdCp`936m9XuA~6$T?*TXsY~*y}082)^ zZr?9$Qn7QYS@TudaOILTK^~oAYqdN+x&Y4VemFQihQt~USi`9mCPqoTSx8e}K{dKw z%|EVn(zZ0BCXvPEM_MLO#9@`ijq@Tlq7c8C33#K1Et-YcFS1WCyf)H$gnYlqFzQ4;22#C13YmM9qf4PQlClNH5}Yw5H&cKxBgwRSxR*2{Q(Ey6!Plr3qR#~}BA z6ybq=2j>nNc3&gDq__v}V_0+vJKG9(gdB?x3&us$S2@;fmZlu@2lrz8QNJ7RU3hh+ zfdEu90PnkU_L6BPWbV5M9E`@E>OFB#CY@Z{@%b{2rs%ZNUpx(2SZvEKGTHFGsdgYk zu_aPbazH{9HZa#RNLRj}Yb}E|jH(H=AB#7B&9<-Viz}BUr}zV->y26-EowyJUOj}z zTcuV?B*Un^5;b5?Hcv}Amn-QtRdTMB$yh1bLMnO(S;iIqnYhvXv8scUqXtDtG{#-d zat={K_gU5F2uTWN7*ERuYa6(zlqE|1VbArgym=#urMla-nNqr2c1;MF;D-h&E%C-Q z9o?_xqgpawL6<~?8ztZo2T^)9^>~vyip1gyD&-!jh%RxEqtGfCJzL(+r>(M_PaYj? z8}&TuCH?yj^*Vp_@rPRj3l=6~ZZbIra$0Hw^ka1)Q6mpg=Q4|kFK=CUmxZf|1Ywb` zJUHn0yP4_(?p%-%M_WE-&#* z7~Jw*8Y2BWs(UAQhgbJYI07s9L}Ru;`S!QLx4#L*0$!iNbEF=UyCL~Wt}62SN8>b| zh9B_1q<|J;CHWk^8c3_AYW#Q-F<2WDe5pHTfR+aA>Lj{?s5&3(qXYUq{?p%uv#--x z7!?GuT8(>*)_({;=gMUXPJj4$bEB2BeHVHBNz`+q1+Cy!bDiwJGQD-WC5yUyk@8eV zZR4wYrSokUDmq)!%*q*CO*(&A=~)AobI=dIYkE2*EZe1>%wZ(Z7TNAEKv{(gDVD-{ z61Q;Wvc!o$Te$%p^kDt3Dy+RkMC~>(WMBylDQn3>w-uCs-YO=7A%`uZC_;* zS1wDe_yfjU8+j@^SRPSzC2<^u__txYn+)2Za?!2xB3&0@)i(=eIa<=KdncmZlU_2R zMDF;i`q~DyzvAys>QF5G0w32bBUBGL40AWQ)u&pdSgg8kc@t-I<+6l{KRR%$zLjI@ zVAD7l9IJG03xyr9UbXIhJ7CBiCs7|3P+wL}wkRuGKc5jdY!7i598Yrf74u3VNx;133Fx7s<0pxf`017L?Ph#2a4v@V9e zVOrrvy7q?lNB(}5v)HjWk_M>908NY!`M+f|!!KBQa9$)m^Lgs_bBN4nT-{F`yt|_d zC}FW`SF~}#WzivOyWy*o=E`LW6n~g60+Jko!^dQ|PQ?&bEHP8DmnF@7E* zm3C=hFyrUcCw$iGiRsEbNgPtB<)i3wEDCC-!T`>_Vj%E>&1Y1cT7afRf_osn?;&8F zjA33pao#W-OT&QH`XokgO>?c!FU=B$MH0n|4qwL5oH>tf44sZ58~y;{&PFpwZ+6N; zh?v?Ub$J7pVZz8PT9c)NQG5}Ms_Z$Vpji2mRV9fwaiK{_$5F`C!hrms0`r`k%hZiOsz z^DWqlC}cud1hyQ4!?^i)g2}-nGZHkgT>8Iw@bf~s^ylCG-Culq;8{ zcWU~>ZELk0M$K8W^o$;wmdHF#-MORECs}%tcF>s!kocKNlQ<4;1W9fQnTI=&ZH4Fj zK4h-McTE~?oQY#?wK<&k?fs%&nH_JEdahiS6UrY{+sJd%-@cuNcI(-(FdxmLY9<`4 z%brf-eneIw5A0Dj->$S?_>W{F<~;hKO}_@nu7t`meiSqedmtE+NcM>jBS`QM(#qXI z#LS#3JAv`WmkX1tBH5AZtE}l2jEzS!fy=5(GdHtl=OdR;-xzZujqjWtZfmX zYsGGO&CD^Fe*2%2P<)L4{@0-L>hFIIZ>V4b&?8MJN7b_+`tt2>1p~&DKOO<|=>B+V z67hE|wc0setUxGWG&Dzs-~-R#l>`A{@N9|n&xVtDemShLC1}K=aRrZ)lshFLhuTMT z^Z#Bw4n%xidG|zgbVFh|tYRB9?kA%~1cf3lz&yXaih=>7>w-bjx`$3;Q852BoXu{g z&z?SIB60O(dZYEV;p9m_zJ3Z8aKnGx)7r)p+*2Mt?jAYu&4bb6lBVLHvzp2uAuewn~-S~^K-i*9- z8>k8`csDwJCN$iFP8n9KD98Jo(ADaB?8%)I@;fc&Ec`O7rX$%D_aa785G`#=7hxmF z)F!?=qj>(^wp~1$tOGS+$~gpV2Ky(97mj?Z1ZhnvB!OZRLcn2+cawV7Vk3WU=Y5Rp zetjE#2DSa+wMhPIUA3K$teCAB zq0Z7DI6|ZW6~B=&k`@FDT^srv+)nZk-PZC7Wy9b9N+AFjX<`--`|3l?akp50scrbN zgQ{cq{n&O;=`rJP@K@Vx=AamB>%5MEzu|~11^o+1wR#;7=q}!dLn|PNELR7xDKKW) zQ-C_|JEq8P46QQ0Av9ue&!gLz(G2RdD`k~}!BD@)(f>m^ghX7D9O7@dSKDgk6o}oZ zYdM374E_(`>tNO+-vnhQIk;y}pr?GBj<&H)wpxp>etH%C6&tU~qe3 z6fBa{YLX}C{!Dc2x{B0zg#?57Wgp@(2Z9bVx*yG^b1|z2gh)E&J)K{`Mi3=Iyys8R zKy%>zQ5douEE%~}&UlBlL|mm{j%oRst`L{4$DXdF(c{rfyWYraY`S3G6P83@*hFQ` z3ZgqO{xkEkk&+G*I4f)SJ_RHWLrS$a#9?7Zd==83A7L-T@%WbIB9uC#4-o1>+3_%o zZf^8mJo?Jat_#BP?Qc_Xjrp)>z;yh~SqPX;GG}>ol*^+hcfGZhlW!92Gu4+S%rtcB zlq%`TqhspV>M832_eb+MVJc=44xg zCNN840gG?3Ayx&HMOWlxtu~o49O2>U@+dGO;FxLUZ~gdo83xw1dx$6= z2HdLoLn66bIi^wWZdH3dw0&8T%MOq4w3_Zypaj6I)+!FGM;G&Xw^UwS#AQM1L|i$y ze<=>kCI$?K=9ROl@DKXk>QV;59j^8g*h4p*TMxq-DDeADeiE2NG-3iM-jZPyTv@@CX<6!vy8(nh$hr_yai`wLI#mI~4RFCJ*P^NSr&HRHa6w)-u?V(&K2i zdzzr&2`6h0pb8gc9?Z<$!LnR9rGc2L`#{GH;S_X9vbS@-;Jw-OH6f{QmSsXx^9NzF zd_D)Y+N+w3iLwvefmV-QBO{xOStOiBm*50AQ`ud2bss*Z63kxE45F8{-_9m-GIbgP z<`u@`K#3kMAafQBgDqb`c6}=WvL8*#w8-B6z{|#F9lg8?9 zRe`Gzc8;IFW<$;wOwJV=YfXZ)8?`PJshYm2w%|Le-NF2}az@htD9sWL)sW`H0obTj zk3ihBul6(Ru%9rB72r+dnfQ3@#6$>g>;;3wlU*Je>(TA<(nH3hJ9@L;$_bKfFzE}Z z{*~ttaz+SR*@{5nDGpT^X9ao7;*&X8pNc5IB$KiX5%{~T!o=NSkC1%7ARCdgUh)7; z7~Bu6wcyeouytJd-UryLv`Km8EW6tex$QX?9+Qm#Z6HMoc zNwHDDi4)3MR#Xy3y@*$-yZ)OIkT=^ELn&8^9!;Js^aznhx6oUJh(ENnNo{-g(X=|_ zTFzJJIOY|nA&oSD>C@+2(P)7mE!9fn{g!V>yjAWi4*g6fftKBD~c zjjux-(ZI6s|HTEE_9vJY?0%XXezvC9H@I?HsuzEB(I)>Z?+%eQ9#2~L?pEMI;tGVg3(u@6+2x zwq<+H>r)VSQiE6YmH1amRNV*apJhw3m7>y4CGb!Pkzq0{ilHPKvgy&Ay-D||fd*>S zuh64Hk9uA`$|~S}j`vCWxAuvks4&A<<j6dwWS~NR#jPSHN`IfgrVjw0lLFW@f1cK8= zilGmpAs(}WX8U5K(F*4~Jsd zSk-0ikOd-vW+DJ0eLAE3EO{$u(o!KmdTEPIoB-px44)3pI$1+~r4-51$x=QQ@e!fmopL zCIKr2li(geZ9+F2p$A0OYS;&o9Q3hBb0Dl60RQ_6a`xq%Oh~&oMA8l@B>ps<(B9a(!mlq#RdhZV8V&EiKxg|B1#fm+Rwi>HG_rUxk4w`3 z+D%{NFP>bHs zPES;AT%}NHQs1X)<^QcVO{R5Jd`hEV?cFED%JZOKO@G|XPPbKXWTyz+gmqBMy9pC_ zsfMce$cb3GKmbo{-%yfF!^%Lk1#mVUsP$^LI;v;}WS`rF|6w%W0&K zz8W!vJwY6Ymww6TUq}D`PZ1wjlwAbrcw2Sm$yEs&fB0s1YqubwoL?LsVOH*vc#u*- zSN_xV8|iXT3TzzVe)^S)hLQRQntf?0NU6eVKK&Jmg^$54h}StbJ$!_+aixTiEw3KL zM*bn9)|Z|YUjoqaN_STXrm#G=oWC8*ysal(5Z~3G^5m)%lKwE!ZlhU{eT;(sPOT@4 z&J(B`Jy;y5V9#*R(CYKLzul^USZ0w2Qf|}wBZx%&7KAb>tS7g5J~B|*NE`ER(`e68 zeHh$RgF*lhmSGXdC3Q=mLZ~?okW-HQ*#^7&w%>#YdRmES@jeuTspPfRc>?h9OqGW} zptB2?baS!zr{j0~b+a#&#;276$4>qJZp=K3o*^ZY|D7S59MlIEi1hkq)C;R6l?RY$ z%T+A-3y6g?XkROx6&W+J#i*L3#g|crp3n(2Q9sdQ5Q@^E`r;ay*C=hj`S+0 zQW@}WP^7=J@d!V5iwO3K6w^FIk&nG1@&Q^JUwL)OpX-CJ3o&zeler#{ljdu8A<$jS z*-7E+RJ}96{gMPgKS7{OxOBt|EcsUN`Y}-8BHI1Y1`1j4+43hYjrw~5Go7yUO?W}+ z!Pz@lJ1sOMi?1OFlP{GTHKIvt43qFPas`^#JFh_;+DQgdIKQQD&DF05;YUcLwtg6W zAdM)Hh#$zLc6Znpj7$W8(%~?JOK$Iw%Le@#Q?%Op6SaohSu7Vn7*2b_t$1=p>Q5DViZB~Ud z*zyN~ZZ~%dDp&gB5Rj6^IL@&@fM%THpjmX-H3mgzp3ZYo`pTk-2I(Q1vLMW5qhlh+ z^q;g2gT}_gdZqDyj4ZUe>Le6bCQh2(q^Ir4KR)HOx4cSEqc~B3l&{KImR{1xK!hg@ zCgO1oIM|0R+l^Y2% z*tmjnU11Snlq_wh1v+URR7G?k_FZ%<)oH8gZ6O>IIa1z;7Hr^wte1?Or&xuR?qBTy z{U!OvPLWD@#@e%T5qbDKs!mJzM$vf_vwp^;TsOc5jkf*38GO1c|H>f^_UcN_&(;S@ z{>yLsL8W+Y+Z7mb*{S=Ra5Z+f3dq9HdZAg0H4)5j_`{ukcCYhV+38=#H!K_O&#6lPhv{`EScg>LmapW;DzV!?%g*ko@5Xn*xp)3dqpRTvL2DLS|V|>CxSB` zh~@dHAsa)jN|ylS`-zT|+X*O{Cw9DD?>iMB#8{s2H#$Q*qX52Q(RmU}3tuW2LygVS z7QksA4I@f^MFGttjU?1EQ2~vqe zH)$HNBaro8FPJfYDHk5wo+z13)S;_#jrbdqwYIhkz=l-yYO*9B+{?#~vHLI3CGl3f zM+J8HFoUn}6FX8*Ty+bo2@SX+!(g+Zg?dF&E@No)&eF3hFwR(RzLn^r;hHBQhFJUt zTdO5YylqbgG=xVL3}`HS_#1t-@GITi9HhwQ!zl|(;L7+gi*IWypz?eiFXGWGxEtrr zULz~qEI453RKS#1{1sq3O(N&xBp$e$|CGFew#HgR@VHrc#%^HS6Z4W#sRHxTLdD+{ zt+iWZPw+NbtU3?amOm$$6omOZM8b=PCu)E>;RdliQy@e~oqCOWtL>p)fI_BV62J0= zkg=tal6Oe^3HyWJF(BvU+ZB-2vy=BI_}tzqT3q~PI>DXSW?^pLMbC)>BjgKi{>dbS z;jcbqeIjn%31qLqEX0$%cRi11D1U1OEd#KnsbL~-$I)n9zX<1<=GRx9jwPda=2}!2 zj^msfsi=!54@{rTN^vrRjxjAq8{b8?f-pm_R;fd5s$B zMozCr1FpSv_OF9@aKk^~`$b@ho9+`P>)#iV-E^hkhhthE6{T4;DY&-;Ufe9xxh0Wu6FP8J*^G(S`~G*6l51Lp1REUNYHX92wqNkd0Yd8<&F z&Xq;pjsvh$E<=_l%Os12C!60U9*b4!kUdIOx3_l-#viGn;xaScI-gO@N?VZWm?rr) zWnJMa=18vAXT?L!h=fi`?1EmH(np~6BU5O$&_*Aqcw7U$nu6kFyrc8GG@u+)cF?4oxQtTC2*YQi-bN3+rvPDlpD>b+$jSx3jff zfcdYGm-j5rhc%|Pr*hl+pa1p$SjCl6REt=0ASl6z}sne3)rju*qbF695CZ0 z7KZK*bPZge1015&9nZNfuq3C*c$tg?q)jKm4Oh{4s&8FT&|n=Z6lITJa@$qC`Iw9B zU8t>!fHjEwvPEsvlbdLBizip*y733|7MjrZvT-q^p-hDE{TO}9DY>E6yMvSvuszKU zau9}hSrla(GT+$K)TlrsMJ+_YCFE~bu$D_4O>eCYPp(K5c$Cs^?-X1$Ki7}m+ZvQx z#6{mF3|Q zVtI(bBafe@Z7h1 zW8@@jy-ySi5}16>r4(MeuAikQdvaA0zDJVjZnL1v{P|V?vRAh;ifu~LW9WWg*|5-R zx*1DfzB`lxD_;$n^~h?aHu=3SLRuDHjz|F3dOZnmb3rAqrn>D~v+gFaGa93~5itbT z3=TmJzC&koTC%Tb?)qv)d2&^v#UJw7ELa_UJY65 za(0$(?0hbQ;>>hFxGJ)gm5UR+5~d=hrS{{51Xl$8DCh-M20aFtXQ-Bh&pk&y*?Nj; z=XcF(d2&?(#vihY4@AMD(g8voVjTiL!x%`FnVdtTWM}KIrbNiCOBpK(2lu5DqGYpa zp><2HmW|gGfaW!IDOCkxZjuK47-)SP+L5HIXM6!yK>#S3$M~&EGv!LlmM4GD<`z$` zO0f7tJe#{a1%*te%fy`m&iy79W08^D9LIV|;VbJ5;?dOYJZMta8v;+!q3{dtlSnY& z;sRk>Tma(kjGmR)DM~X340Z*i5uNA(n2t8_5-tHN5XL?n&&e*m>+4`X4L=4Yr@XC} zA2~Rprd0ts7}t>agGVs}D46m(DkEmUw?D1USvW-smYepMTi4b_)LPLm=PgtxI~~?F1}#T3kI|0eU&Xk(WB3(dsY8F10;qX zyqC3o0&S-%!`X&EV6?TnQ-Gys-81U;kzFj5t_44uI5|;-WFbm6K|zpF>(IS)=Uwn##6zt5T!*RJvxTdNXle zyEM^MDA#&C$bb~#AaE3Hox)&dpt-n+ZzbxY8Enkxgc#A}HoRqT=gX#mIU*IvnX(6E zgZ!gO3RAsukYS-H?C0#fqjpB96Lfyg zNgv^?4vlLXE*2z-vlleA6X{&`=Fx|eVdKellq69F*HLRun*QL`_HL&jN$hC|QeUyO zYKPG%oIcvBv)ZwL3ubO5A{j0v@bGxNP8*4W zb+j1Y+J3(VcZqi2n8gEv<_b^IfGHS`xouQ|oFW`H{Q}>I9!0D^wp#*baaxB8r!$;H z^&!M%tQ$#ZFO;j;tH>$gf;v|pI0#0!=%cdFIlaXO&Ymp4javUxd>zaGX)hj0MxYZ}?vj-T&op z{+|D=4K&yKhd z52?aFPQ@y9ZFFsHEe7k1ao0v;|KneTzv(?J(uK=A5 z_%9Hnl#cYNQPx)l-czL7@C9a#R4Hd+a%`X;tN-@JOa@n7;t4PTigGMWPWRjmwtfG+^#_5qyXw zEfe13)ujl7a86{uOzsw3D=a@~;=A}HYLWRD<335Pb035yMT$D!_7r$>MQ#qizs`c>CIn_V2$%pKK= ziu?^TJj9QUgr*+4_9g;u+XL<=rW9`0KC@U??$W}b#1o#vldBRW{=io>&^DK>56z70 zh1bJiE)Lp7cu#zoq^fDrz)}1?VT`x5zMI-*`c;x)a?fJah!#e764N@;h3=t*?4@~* zt|5HM)IZw>zgWz>jzA2 zN}uwW|3UiTC${5~il?Q^Px*VuWcQtuLx86-aiF6Yhu%enqXk{EzE82DW;kYVj!&ga z5E9UGDF>cCJXr1|7J6aXpZ@W$x}IVqj__xhq{CTnKfzU|eEXxr^mjkBi~Ok56Y9yy z3D-~af^@B?8@Jp=23makAn{kC>w$%Q5_42zc zv9=PGT{aLxC|gx`5Q)SiiGkpBzj!#L>#1G5kVvca%Sr}X+fUg_ptY+|wkpv4jRJSJ z3j*yRuu}Cv?4gH9+EAS2B1P-@4Dp;T3wMyk0E3(u>Z#Q5ShphE*iEChNBR$TNw}K^ z&2LpHB_qOc7*)<^@25C+IvmVlW~dO3#_ zZ#k$gzq-YW@}7$u(M+8Ikc$LL(JyB3lO`9%mF~2Cs#Pwyl`SUo52N%zj4T#-W7-`H@ssP=Il=*YsqOhCzHOQx9mb~8jUT}B`v zV#uS1moI5gLftqjV=od+a_#Ps$rQQ|8OuFug(ys=lt=_A5ji8&a;azAli~w?VZXCyeZZO%U7Y|Ik<|d%Pc@sx-pfRPS+N(#|CN<5ZY1-ta z{#tS%X~8T|?V3!Y2s%=KV?r}m%0{}JO2$XSSFy6)s7kToZ{oMT)hJ-29EG1)#}(^H z(`_!FueUZ$ZuL;+0(VQ4d`HOyj=T&ilt6vv61FE-n!-iJ724#w5ta2e-`cnOh++y+sB%$=6_sDjXCDT%IA0@K;@c!9Y32j#7UL*!`SQ9n!>C&A(-9%0{@MmZ$6)<*(<2vt>}^re z{FcNLys+!*IEX(=xuE3joAFjTDGN@~ zlaz%$xgwX;wm*b;yKsHy>@d|kv3;y%YMzv0`ffZZpr*cu;uXSY$}4L$ehhpU5DsP` z#se$>NupM2d zzBaczqhBeP;u}prq@>msPp--<><>=eZWWC>;sX%aIo))d_$yJF3e?aPX zr%|wqdSrpWK5jV|auuh*=C5qrxgLRAA^n?hb9w+r{IFGTVl zPCBMa%W8j4d`Vmg;DjU{tH!gbi$JWb+B;J`g%fDc{39pCjv-ui140#A<)Q?m+`nG! zeOHw5YXR!m#v!~?JhiG6~UN|8}40f`O7tPjUW@`rygJSHewowfuqkj>_MD{36 z`)W3yPm?m*B`P-n3KRxoVP zw=?!+MlwwPx;OB)chB3__SGHoT zL9FE7-7D<7;9XgzhmYw}LoDNlORaPSvE#`F$68XSXVW;1B6Tf1zN|{VwJ$+S{PpA$5mJX@ z;m9_8hdOdD;@@D5GHS`d+3__1Z|+oOgwXKE3vKND)k?s2Nz6=zb#P#ysn%H^b0g-? zdX&XRh+6M)SWhVv?4@i7AFT&4;k4{QUs*>AXyh(@^_C#cX`1OLjt$C=3AI3UgRDoY z8Ks@++M%{GwSv#cm0e9XHj`QKMc<%19R{`c(PZcvj?-iiP>N2eJ1I%YT-<G)4K^nCUL51fQe%2ADJ-7m5l;C%)v2j~S>VBLgjK5?&J? z#}9q30bQePh6TXE^YlN zbr{Pl{&3S~t7yiO$&y*DK=5L+Vrx{oo9M9@;pavDEvA$F4U2}ylK-TguaWm$NYPZK zb*+rgu> zcC%BY@P9d$4Tz28kb=;BP6{EgOq0p5c0zGe3$CSA@^z@>uchN;6k{Ql+;huFL1;lV zhFB5BQtl_K0JtNkhZ!>7G|4sU{;z`@_VwO{v+FP<*&K(ndxT_O|M4%QxMbtr@}v$y zzNv&dgykE5NNf`gjLmh-K^R@uUcGuH;|B;gqc8*SNI<#N@y&2}#eBOa^h0k@9M;4PO%_idb44E%kGu@K-<`k+Wo{Y{eM)LD0Tq(fMWDrYo=}DgMdxB3Z?CBcY!BYC3yUS z#dr`E1kan;IKf(Z5hHlOmyn45s)pK0?X_HB!aIBq`^m=`y>1jRIda@z@O$ZvM#1cp zd`|n&3)rfhe+tmsJg1$WqZO~bgM6l%CRxPD8)jU412uvF8VSsJdM2W&WW;oQu}RvU zs<25sN_=ZjSCDRcM?LN>hy3<3v9m1y^pcV3hpvc7nhj)9S1!?H*o(8R$`v+rJ@nSf zE#j4YSV=jPgIEer$;I%A`CQ5P==wtA+TAJ`YpaTF`onQsTSeE=n@~y;0I@Da=CvM- zWW-WWA(~>(swD)h(RB$OvkhakW}8|c^73moc?z;(_tQoA?RD>C+21YJ?eDm2YKIBp zCC6bn=ZDOpBneyw?r-ltfk5zG7#vTo%EjXkv~6uS3Z!>++00Qmn1?4<<=p*YugFdnG`V~1 zkPHX2$ABE{ixfB6M+tXZ!Cs!p@{Nb;y##Vuu2%LdvrgcdmbFPzx z7Wdyyf2wH@au8c%@%H*7doUE3unS23#|folJ^iMw;c6w9Y*IvTwHD!}dnXRo(B@8mH0P#m;WrS6OU2mnuZdFPPk3!rn+K--rYgLeb=Ybai<95-iM(v3$*5tN_b7@Oi3hsjWNq7 z7FZSmqqqG@NN&$StDN+g!en)w*jVKisD(P!I;+zmV*Ttntx1RF$M^J}=>Sf5hq{hT zAMDD(7e8yg(wIZfrxnJzH=Y23ms@|y4%XxQ-Q6m{5==d3F7X`dyER@YFY`KI^qW^a zpuq_*G&FWxd^G3;K#&H)Lh~#E2oP*E3Pt8zD@$m&MX5|RzgbcXp95<8x!~=|6?x_T zQQ2GFqVYN!#|*xm9M|f-J!8%^mog-mf5cj1whg0&39RfN1{Ysd04nYfLQ+-d*q%~k zOT|qPx=zu9O#ji%F5HqtxpskLe2O&j^Sr?xhXp&Wf?oaIY5nx_LLFObBJNN?TPSe7 zc^$UGDA81V!0Wc|Twz0k^-}(?HuC^V!gnB2@p7Ab8VFN*c)vPzMtlEXI}3vY%_&n= z`Z8D#r|8M2$(~%1*WBX}VW(*4&7!6DikEuowjUQY^e|Fxrd{^$sPn~wn+M>AD&EpK z%YGr1i`|3?e6t>8I+?Y9kD5fWWo9X2AZiWUv6@a8?990Pc^X5Gea8^Y(&bf|$5YYAMWYJQ_RB#R&j%m+g)?k`GBtT{nX6ghixMP6lpBZc;M5r=<| zH+eLZz+Rox48aZ@yJf(_N9~dw5TS_i2%}gH8ZA=nyQ}u19~O2Nri7tJf)`&XvT9Rt zD48(9b<*En&O+Y)XfnSJ*h^!BNK*55W0Xw%Poc*>&-?CCYq#AfT7qBw+c$zE$j@p5 zY>Ug8oegg^uQhu%q=4KzO+oh9@hY6SxZ!`!J7d0aU^C5ALHvK`o zKxIDWeTQ8y&`HAws7u#eM386u+#mOpnHag2e8+cvH3{3hKYDp1gB?9;>9(6Y1ziH? zg~b_B0pg&fuk)MJ2;jFlcQ?S>DVl1?_G~$FeZJ%aZLGsmBD0WA^cU0nCyl75d|zrm zs}1Js3sdCDRmma#Kt&b;3XXkm7CCVO<1x=ud|?>IVzSBpTWT4r#_(}LY3$e9TCTy; z9TW1DsftwF=n8ak2s!`IMPX~ZA+Njw!$U@s4PXH1~tyZV7-E*=Ct={^!d zDT5ZMF;V6wp~&Hgt1II-u=taTK$A80-@M6Z8Cnf+iD|+vs1`8T#z6d${R|#3le90n z2Q-^jMU;k8JZX-y?T06APp(SR@dt6X+wFqFMdB8>l$ZmZ)!?QMQfU<|D(s7Rfqg`( zqYJ&L9exhUQ;?!^sm87&D@ho58EPr}US^W0|Efa~gI4rN z$`jQs$Ly|aJJ5A7+Db21t28{Vpc6kodLf#O& ziN%aX<~0vswF16X;*c_ky>zs41f4df#<9Qx%ZQU10J24ui%cg4?!bS`Qd_w}1H&iYlwv@CU=P;Zks!y$|lgy2+{hGL1eg zm0$F&3F$}u_@h2b)z508`)KZ)rLSE0nA{+XKipG1_!B0>jOO_QGe;*=zON>{{yt#W zbi}9RV1E{t1ZWty1|%GS)YX_&fR9Z>c)RcXPgusb0o z-!o~;01F3jNz-0#fCrCiB7Xqu$;AW5O9p0jdq!?OmsdZbp?uyobS}L-ZQRz zboXn?OR|i@P_5rkf(8)7MA4=lj+AGFJYX0;y#E3kKUc@3UDZc$1;A3`VoiFcTn=bGMY-yE zYJ89SxgGWcH!njy%tnt;VW_`M#TZu2*8{tO_v{qWLL*GeRp>TMb-<-!w#x8pKsLJZ zX!*=vo=;eR>w|MN>ecO9Bi}Ofs2w2x9+XvOsw@3a?CRwp+A(5BP^)K-u2Ps z$vQBk4S+(1^*-L@a7O@KKCKNGI73c0Keo&r(9iA=GTb+jTDDFE-g~;(D5oD#IKHHh zF8O|WtI=E4snPvGah-0r0B-q(Eu%nWhcjNn4>6jf>st=#e2GzjmY*a5hPTr&y0_rS z#RIRXy_LMUROq2o@@E0Y6#d~iq!pPon==H?svC1{K1HYLnciHFkl0SMU~TyT zMp<~gyk@ri6J4?X{F*mxM`fRMoRx1k-EbPik1)pu(ZBsSswIjgK=tsR!rQ*VKq&h6 ze-3K@{@=sVJ>Mz$@_I6C=Hz#Ov@+kSVDgB(?G`P!tqD%UFaN;g7dje@5$7GB0dPm) za|r@l4}_)Mn|ov(5~*AKLZK|`v45qF;#y79yM0r8Hni6z@7cB|*`)RsPp-%q*rQ%- zXSZ8$!LO-!)>9gQo4(bLBr%;%S4@s?<7g$hrVzWKle&lzhsI3>I=b-oJIKks>2MYX zv{e)HQhA+#(jjv*MT!ZIG5{bL-8F6XArl>F&XLg`dowhY`Z05ua`~g(^wn_jhJY)4XCw5kx5Rdfv@cF&)h&8UooI)JDw>_;-wy5+`mYIBh_ zp5j0KJk>+fAIjF9f%^!r} z-3pU^BnPb}wrEOD&{|5!%;$kJpGyZl_E(U2zKCKpecwcr5KZ6SIQhtJ@e1``vz31{ zj39^nUG0~}O(H_cfcG6Xhh!37!36=e13MT4@egj|lI>&L8)u#;S0#S@L3Z6{k#5GT zmq*5FRs<~&ZFPKj;GsJX?t$`bX{CRoDL(sX&R^Pu_#XXQ@r-bFyD+Y&HvKuG{vfYN zl_7K4xd_I@98vQ+5ahH+(HqL+f>A*ElJzMXOV4yg_#+T@+sz_A&c&-UG|_sz^`EE& z0CqHfOJx=15>H8Zoy-^J_iWqKx3^69aTb{+cVg+}nSwoPiABs7h2-QV>&@vT?9l{6 zPK?@gf<@mg)MhTDFN|bv|M-__2zjd?F2SS_QDlcobbxtqA0eLdcDqI5-xqR%RPPD;tDTvY$h8uT&wie}F)o9fbl;Ahv z;!`}k$=4SzXVD1CA@#+}7vzcAa;`oD$m&Ml3gg)bG$Wz@&^XG}nDviT>G)a-ZYL@r zXoHC#Ff6PcDJz3fa1-(Q-Za9Fq&KC1;X(>ZQph_y-tM8Ds@y{!rCfIk)f!fVObeWK zS-l$T35Oyj%5~8-A#U%Una3>pTswqIplul{6j#b#oAZQRgYmo{P}91NsoVu+97JW8s9yov~o2ns&<|-5l^m2Ci1B5y0cy6?#k{h zGzP}f^BV__KOh^A1M6TpMYU&8GmU8Hy*7)k6M>^=;cYxy=Az^^c52*N=eG62#sMty zK^;AY+B+^NnkeA6+Sv8{3C*0^!nYBovT?dj0DcNO_QG$4HbN|knN}2Cz zia8>T!Mi8m*4+(a5lGo~*GjI#u>SVL2e5-FP!3g}%dGWFsM=V!@a7crc}>?~KT*DmYQEvMM@Lk_nmS0q0C4JO*#1#^iN zwet{7C3Xe*0E`pq)6)-t! zo5>5x8gd-t=D>{7Wc)SDBAkGq2f~|OaPh-&{pcAAJkIzj(yiZN9cE9iNc8v{OSGE> zq_&lQZ`wsZ7KL@>VFy~QZ5iaM3hNA+fA$x%KmJv+9ZafTGhm=u1ltZL0=231Tu7WI z>_3xEUIJ80qkAa1=~#MC*8rd;OAJqI@5D$Y*4}jk`5SudwhKsgr_p)+XcB)~@oi(=oco?!fa33t7LR#K=qf%$WmN98Tk1=nlQ&JB-C(Xhjzr1vK z3j?cnGk4h~8ELVUywgJg(g<6K{VDID#SzUXC_J4>r!y1 z_-ihg{p^s@vgfxxNJ9zC>pC!IUL2c z<^i|P9eZ))8HZu@=8RRBYQ;9JMp5B7dKKsc$8I%^1IxkOrAf$l`4Ke_Bpv>8YJFBY z*oyOdicsu#B@Lcjk*dbu)CG~&BD$WAd|-U7441N+_YTVf8@Oo2Cc{a{5l%lW7cW^j z zkm6HI0w|I{xSm)+YHd-$=?gP+$XFk@1I0nsGWW;~HkNyF`9gcwr4VgS=qZ}A-I^N({sct`Y6k1gKlZWUZV*7vF>B3wQ&KDkw|L*v1z(e2XC*m%utib5V+ zNrbsNWl!?hVv?sHFNiT<-x0)7htdJvvD=w7*;m>xEZ5E;O6z5yMr>3(WteMGK@l7(BxL#Rdp@znBiwh#- zD)^}9v*vFZBylx~tt`1{d8TF44(hmm2q5<4iWDf@{y<%Xq6)fR21@(!Po$6Aa0qWF zCm2OYWJj;G!X+_YfY0)uVv&Nlxm|?Rn~_hi_rTe2=E~ZugrLH!JH4jrDs*UhxR!J! z1@V|Nmx5L;kh|FUKA+*gS;9BjVpZ*fP{>r3uVgLD<)22=k5mYe)vSV4h~*!DfG@I% z1@ol+xW9k;6A=+p___-RVp<=F4j$p-%pw79?4ZDP$s}NEQkq2KaxcO@WtxZWPTd^9 zB^U+9X9Y(Jq*$G-01?UD8fmU5+Uv4Ka^oo*N#9i@^W>`JAAgW*{V(-cFg+KhY zxm~0{dPG^mqJeHB(U#(@aif+jB3Wj2oo*|+$nf$RZbxLBc(ERLSI}g z*!97j$S#>vfL0b$BuWYa372{GMpvQ<)pZAMDy}6z$fZ#uo?7hH(Cv)LI7* zmY_K}O}N`JHA?*mtls>CGic7W*xCMEEYZ zDEac&1kxeHMh&cw_z63*v`Kaml?)(nEm5O7Es;O8w%OP&Xo=p!FU8%(j3pN=qt$+y z(-l*dpl%8DJ^@}8b3uX);+&$HDM~7FV05=E>Ez-@CCEp$#Oiyr7ESR2i_K7#i)096 z4kQKU=?xq$l)PDh!&WSy+Tr&~@eMm$O;4U7_T;KmCH`>RW@D$ID>AIkEyswtlBpqP z5n$uy%TR-ynIo%apGZ3)3!Ge>oSGqZIcH%G#v8ebCs*XiJ*wt5+eN|)4{*sf3l1Um z!4W$PAzRP9&B9H3gWoYUcD!{`!uonP3QTwLoiuipFS01foxiS;4}wVSg5jP~&#T^q zt3!gA{J-5HvW+Zve>B@qP`CcBW`!qLB}V+gtxcr4HbZLr0H&-|e_*Ft8~3(feNT<5 zQ_nJhgY{`*Rfb5U;d&C*p|Q*Y%*l*;%ZD$*x|D&W`!Lv)D}IHut^(ES3nVv;kIUDp zGWa|ZX@7s2#A4V|cxVAzpooxvOd9w=+W#D>{jo+^<3u+cc*FN(p?y_0&bq#sD9vtF zm?#Z@&@CJM1<=Vsum6*lqvqiR4R6*GAZ`zxiK)(Bk!2Ohw{iPZywPcdEmCAEB%mY< z3kSHVa?PMOEFMzkoD$Uy!GgPx!Y$UPgtnASN;`fQKe-Zjs<8N}zPaHK!ENpqB_+FV zoVQDFGWn@?sEU&)dC>C&ZIib(G&(0Hv$mQAAcD#91b-nhSWI8?Taph}+NMyOAuHk} zq)42uHnmRNW6m~_EK@FITAmC&Iyn4Qg|yP}2hBEjcMHabJ;Y+B^;h#4!1YcGAv0F| zH6k)iU)KKFA!lQ17Jpu*;L8)z&6$Xh;RKh6*$0Jf&Fsy_CZ@LD@xxY2~%L-bg2UIS8F{!Py*PH?Eu(}>QGfDNu8BRe&Cp? zFWu4bpEi_ZyC=O3Fv}h3XYMkf2>ry1?`SE4)Y9*tnj~?0$cTvRpREej<~2NwleY?X zv`EspI7H7TqrLG|r&L$I8fmqCu?~20Rq};DjJCDCT_8h&yd}Nz`g)jX)L(I*MCK>T zV97r0e&_n--<(FFi=l(y9);Z_)^y_p(yTNAP`~_CdrO#=Bqz1HeuV6(GP;$iGW?;f z$m|znqF?J)nufEreGVJuu`4PELTAORg>T1|MG(aQQ;`nm9+=_RckOT3B% zG9O1-Ytxrl$wTJR6{Q?UIYO)K(vy%{X`@@?jG(1h1bpDPNUe@#Cn$4_l8s2m?~9gD z^48h)#b0UdRzbkTx*vZ?ZL7Ie0O|KnrKR~OzX**v3qu&#r`$+MFam$MHCs*Va@~FGq>J&*8 z0K=Pjs$PCnk+c{F3Q*k+wn~}$DK1|2-ne@Q*Fq?yT8 zm7yp#{XwnmW~;z*>HM@jc5@QT<&vMBCgTiyy9S1H2p;rW5vul+k|lAc?Q1^B zldBRG{y^S#d$(Z2{k7pKS&1tyZHoPW-2k~9az>jY0=GeOuF|1Vf*6}`kj}>pr0;5H zfBdVg3DP!MV9W_hk+kW=6p=&%K`>4o_B0BCjFWJ1gQ91-2CCEbR{wM>Q~&rwYv~e; zrpd6cp+a+@?0FgUT6YjTrf~uxI;WM91y!!r=~IFb8nts2edps(!Q>{ml~L6YbLlfE zo52v+zIV;68qKDcm|>LE6YcDHD@S&!QjU1ki|%Z17qHwodZ(R4Hqh!HtP@onE6=M( zw#MipEC%D0jpE>Z2niVG-UpLhfG~;{)6KDwY#CV;ea)4v-d!zmnCY2l2IZ1Pcgq*C zjVD*+YVoKDjd@3b;N&aEkX@=NI!0m*(h~5R)-fb>QNNU(P|Av2lY#yku>BRDHP?hx zz_~iKHSCI0+MH?KMt>R|3UG<*$1uN%91;r6(I*wM)<%_#kcO{{wA-jkMe0$Hy2EPZ z=8E*fR&;>a9(6L(7F84d%XFTwdVEjN3XGHl-S{a}1@bO#a{$xBZZn?8z004}T+_&Q4M9@`jM;VY2oxwb6k- z4v_h}NXlwKeLud}rn)u-8R`{L>kg&CbA@j!y~+`2?v3kjQ!)Wd%M(K9m_>|oOJN*i z(+`Eoo?MXt@i*`3>^2G*OugF6<&1)Kz-oCai_^Qio5jIkY&T4Qd3)=WmK@yAP1Fx> z<&TFLQ-l*&D!^3G$*3@qZhq~}C|Q;p`f*ZiKejgZcaHeT*0m)g^LIC4YtK4C)xww z8cn5s;VRjFxJWzS$CuNpddbx&y%`p7fFUSYPeSNxuaNg5(afjueRRVdC!T9IwwTJ4}$V?VxF{oRnI;j!eqpG$y(;{haK8}fxliF*V;dyumZqLjQD&@HLMqH*~ zo-jw0r*ApZ8+wLYewdH;3Tfv**<_)gpHY;9R08HbZ`aI4IbqVB)pCdsTg<>bE(LikUsFMk8A$68n) zcpd-of1~LZpsf+jcpUxvNA`yJ<>AK=qHXkV|E=6Lwfz(WuJ2&K5mt=~a4GGY@;Bw_ zc8jR|EVY|+OU7yev32x?&ytb6pW8aJSy5A|p!hc9(}){m8M)CbQb;T+qA}o_5%2p% z6B==-rb8m22OEookc521Ey4FAd0XS?5n}H4{oVF%@& zrhsSf4qnQvoE44_5x2})GFe3RcM%&pG`_c?3`4DBI)ElZh&&de&D}*QPEXxT@yaCl zXMkHaHAbeFL2+wviwQ`U42HHJ;GQsRR{*$ou-n$3{&Mj4mHrL?7yfNU_ka1D>|ffV zHsC^^|HGfy4O~pZKmFzJuV?%;sEudghkv*WfKKOs4hMt#e{5_u8h>`*uqyV?~kBXH9Tn*sfR4r1u z6gW*`MqOHM?PUMMhJRVDZ~pB!|7WB>+-@@B+uH8>0qO0@6{$$J{n79{MYeLTBEcvH z&uK@AQ1eX0BDi(7)pE}W8qvg@sM-L1gVj!PT<;Mi_apj<>FW6n~o;!oQ=}-0oO@MRT7#1GL3F~(+1ltD?k*Hb1k#M(7 zi(d$spirQ@$V6hDY%Cteu zwS1K|2Zb?IGrGDWGA1WG3qjF~1giyy1FV>!1l@RiW*5W2u&=DWJJ?Jr64i-WE zEi5%;0|D5hTN&qGgjB$3BitJX5EN}2NY|C*&7Y8oNIIk<=LPm!db3uv&@D$+=n`Or z57L1A;ODYAX2%Z|mH^qQ02S6%yZ(^c?PifZ%h}}-Yv~fla9|0W=I{g9`mqi4d-|Al zgDBNCP6%HW4Zx}VMZd0NbnGU>Bj>q!6(g)l{y!K6SOVmJj z>rL53j^vL!n+22WQk+@N_9+V1t1kUdt0VDeLHTj!#9gruxO+iPH&tYS{`5)x>$Iiqo;ohh{ z2Xz2VelP2OUCyZj3OFuI{K2Bz-Q5Bt?D8#V4hAM@^fW+CI1K?^3v@0P8NbLySgPX& zuyX-iN~&TJ-#Pd!zZG%D>L_Vl@Avs?ONRD^SBJ1rajFzxFFg#_a(!&G=?m%L$rULM zJZeqjSyjL(!RzML0zd=A3&#_DxX8PxaSgL4%#`>D?z(jKTZ6%|o?F;>Awo~-uOLX_Qe}UY) zwm*BzxJmXOhI9+#HmIBb%usIKEudP@bLlkv0jZ5fQJ0I_FNRllsut_3C^-mGBQZck zQIbZ*Q*d``MJp0g($UWf2^AH{WNGjBbtGp=0^+59VE9<6JqhlDjxX?oCs!q#_`^+c z2P#;qa`ujDE`^CPQ;$r`72rc50r%sBw8kk&3Z?ck=9z+k)|8@4>WfMZHmXCTF{XU< z93DHqz>MZjRlp37`p}JTqW~Ox0#=c__)(Rt?xEZr($lP-bU=I zGYl#<*vo2>vmhp;q|SYp!J^LYp3JzClI=xn%NN&`Cs!qv_`^{#$SFuA@H@iBMI7Hs zLN-^IS|ytbl{h0MLsXg6d^NXYS`R+)Mw<~mTDeI85JGG`Dos(lp-&ofdFg|H{9h>R zm8&~ip3HE%NbRQg%a5-+{Gp-ER*^;DWsHQ)aBe$z^%c?m=OmY>iBFb*h628?j$W~5|$-;enL84i}ag_y%3kVcWR z$HgM*n2A(BlgThA=gBw6o?Mlf@P~CaJDmbB{Cp-ydD_-AKkyo3;n};MFJ@_C5Z;L= z!JPy(-A2t`utbcJ_a5(@aHje-7YtX48IuqV)E@LIsbF|Nx=WCOP-pvc7LH<)RR4vh zZm2XWkCeP*y51^}Ze=PDe~4sk73>%|WVJdm& zOAeUqU})GI?hAG@O&H6$@o9|)16UQQbiI5u-R{bbO{29r3!$pJ^Rfj;mR6$`YQp(PkK+1-#wb&^WjC*0}vV{|V~qws>J1QF&zZz9(s z+hQAP&Q8jYpoeNXpGyORbL|2pppd;x6bnygs5_s%z@mtGOUc1_yY&Q^`}NI2P(D^l2d6hv;dn+4e7c|2L(Q8+~dUT7mTWE&t3 z9?1GQ2M{Qn>OIrLb%TnBO@U(8-Ek1ZoD=G30Fp?VO4kYeY(CT}6z z&UPD$9&e%>f@&6x2>;cx;~84r>_(O>BX@l@neE-GG?`6*NM@_OQxHuS4%pTi>+1Bp zZhy7hv=%6jk?-8~#cu07u{0BNi>YlKjh8f@PNQ8SF9Sr%w8$6+#FaiC-K0N#Mg17p zPCgMr<>CWLJx_cco?Mk|;}5%Rb&59EULo27hamdmx(Si)dXrR%!NR0BG>0k?4IpZK zUL-S|GK;_>+C+^8u%_f44zz(eBx2NxDA}Ud4nuNHtIA6iwqlCp&&~_}cDl7lcorCIry@tLa6)SIvVO zqHrr7Y|1H0o{7@FRIO5`@Ys2pL~;3gv0dN(M}K3B&el%R^`GrsZ3dRioXh4))cm&8 zXM5*X>{~L2m=t~Z%E3a#y`~f=TCjSQwN>&ur5FuZ5Di@pTO z($A*1GLxisqK&#qGWRLBtJ|THLGw9D2)8cUC?2;Rp-(lok#encS$X)Fqn`5k+Trw}3PKE>!|gYkY2900!RJM^K`g#uQ_IFo2}XOofmb znoA-rf1Nv1n`5?1kxnwj-!t%^fII5b5YcTES>No<7-^YU%^6pEs$3!}q(myv$rO?71G;aN{80^We9r)pItiNGIw~!^{$S%;R zu#s?B@q;L;Rrf4HXt}rkpaz5O2xMrf$D9j>W?#ZtrF~T)%~nC3ej2ICE<#8@&=I(`I%2$W7LAoRnN%Co6$5CbN&%YB&IcE#p`zLyZZsJ5#LTzdPS8nP_&O5Ms|$U2b~>6^r=T z4;R5D8m25up(kfJkfi|LTZuosk0!(1ab2O~p{`KpA{k0)ni{uSJ5d%jFz)%8F=pMz z%oEdiDxx}E@k=JSY^2$8XGBBng}Jt~#gnU&2K+&VJ4G0janNY~C>+B1Py8XhTVAR) z%Fy*CL+9 z8!x%h--JHH`A-H;!k>UrjtHmd0@&ZBhN^^Oqs{V}py7!JN0e292dCO}+aJ`q-P|eA zt$KOd>-Sd-89Eg6dZ$9*Cb8;6n1clxZ7zjL{|<5#5s?d$i#VdTn2P4WCNQfxFc~6J z+TYdQ3c>14Qy}xzR79N9;GWRv-9|GmtC>pF#TROm7koA4;u@E%Od3yN!hKg|!jr3# zlKf$u+qgI}p{XA%0a?$vnY4oKj}4ouZRyNr?jZ*@MXYscd+7V%(8SD#5t3T;dPL z+}>>$aPi6c=8LFfqo?K8>ZOYw>vPk0@aZ%h%|3iS3+n`lQyqEi#raSStyD&=XD&!W zaQDY76Jj$Y0RilYxa`ukpeuIN1iUfSmtuf%u+BxaxvL#R)qcUqWk#~noa`Acs~tbf z9t7A<6_!0L!2H3njZV?BM@s&EJVSYpHC1o6P-2>)I8;x;+0_~`!0hS{v()pAM(9_# zd_;g1)29lb)GA<&_N(FhC%Pu zAHsMR((vT>ZT(-n%{z1!&vXD!VX!>U0eBRgZ8f(FW`+mLX`n*vH2#GA{A!&WHW4#} ziJM929QNrO>xUmcL<6HAuC8(H z(88nJ_NxT?P;s%?ZwhawE0T+ObZwFqv|ZdGp3hCht5z11ox z>i1*Ao^8BfNg-xmGDsz?*NUu6C$JO{ZB={9Z?wE4?G~tPC9}y>Xe-Zj75GC6+l7wf zAa4Kjzy8xg3M?Ze=+7ihKNvta!mf3XEymGI6jUpFIXkBzY;XTXcpKl73MAFGI$#5U z>gmcCiz`qcpsV+1k8uD;Q8?MzbI^Ifjz)@)Y*VGhGt?z`4#&vw+jwf4SG=bbfmHm;JK$ z#{TW^1rwbP)Pfc0zdl;-r66uJSpw#(iFiDxTs!{UHE|Fa)|P4j_s8=7h0?a9M>yx7 z83p5Ma1(P$Q5@a}H_0vQGTapU5dDYdmwJF%Tl{hx+y&Uy^Lx5vbg7g)?)DQz9iQpA z{Xu0g7YoX^a}g2O$}`6zk#IBbi>)(%W;_G6;K{xE70%!2v}q2=>}D9;&MDxd`U%y7F^kq=mCBOy!c`${`;HpjUlU6ZTzC9?4ppHg9+G1k$SO6Bge?@8n# z&lD4x=fG^FM3gWX0j4SPk_T%ejvs|QW2MSdGPrhqVGFI@$`l6vAi<)tV817Z(ggeY zl%B*AoI!(5c_&RYPjEuKl8&GS#6iSfgCfIP!4!^Sy5rDzH{`J^g^NNTaxNzbiKny! zyM2xknf`tSo*kMlU;GxHT#>fIqX6tqK~vFREYaFI4knXipw&leiUXB{G@Ns(e@+L> zhlWCTRI1DH67MzkXreDFLY@*$&e&H|8|_Kba;D|T8@bBso2ZH=JnEWwdKulaLq~BD zmJOPwFNSQpSp|zt&2v18(QY>iK<|F|J0ORtQu5bKgm&%LvU0|e^H&SD>-)L6TaA*% z_=|E+yIMJ|W{BihAm%eo``WE1&>Fw1HRZ`wNh$uY%I#KBbdVP;ab#xCn3;$bt%gDn zm&+Ztk#`a@E1sm6lKtA($_%+!0hLfG8d3P*mS!{hxC=aYk2#Sm5)EbXCh287JZ2<@-yDescprkgwmtcrod%F$93~u z>BJk|q+J=hNz)(j+1x5n5HT+-pd~HeJl3jx%GClH$*vYhn-!VlGt?xhHM}H)WO;-T zh>-$Vef=}~v*Zqo4YRyBE{SurpJLnmeRs`n70fHF67dIvHjCPw-a5FF&b|&<+L8FD zK`NVSQwPF00Ci&GAnAN1HQmbHID1Ak0vvPgu(L;^*U^oW4;*c z$`vJdC3WhOW&ag=2EUz#S-H-rv+b*l;mK7gWBft2g#y{Xo`sn8opMl6u4lKY1MEuG z@ zT>5M^od|q}xooHHs~6_URfz|GU~ad_!?AI2ZmSy&H1TIqJe${g?D+wm6aXVN8>}i! z*4ByhleS+I`7a2lWzC#I)?|d>>oD!hdL4b0{rw_|ZtW|(lbG;0oBDlm_SG5@Q6=vf zPY%RMQc373~LzH>4MNF zgW5aFm6vhM6!$2kR*nWSs~5M|P>9PV3Jf_tF-CcEMQ$07vbVd<-OYqihoz!td6PCg zpd_xtIYtz%{xg`p$lp(gTu-SzPw0Y=>3X%T;|zgW17xd^Br-(_b~T zyiYC3au*M{t4SoK-EBrYRE5HZph@?inJEW_h!q}!)yQu`RlsFQJDEFFnPA!dU!|A3 zQ*U_G%iV777Ayk0qHv`_vkq?nxZkdjHGc@V_9s z|I6R}J^xu7Xej;eOUWQ;K0%@MJ8(~epqbig{7{V|1%bz*NxNtQ`#xAA zMJoAPF6%{rXtr3vQ&iBjjO5CIOer5c;sLD6W%bul$U65~$mK2ws`i`BO)Q%uUNTUF zVU4K45%#x!v{dxCMWn0b#zy&M5+jFTrVxc=YfP-Pz!@BC&~Whu-2rHYC1a`WXP=i? zYNz|WKRlNF%~!Syz1u8<9KHNSd*YYzU40mT0;;;Zo5cafJy6Rfc3fx2kUxen)kwJE z3_F%0gd-PXqIX!x7@723GP~SEhEyPeyc*Uj<%DLpK&;uUIL$`7wp4V$NFV9GDj7tc zY~~4~O4!URA^DrbY zK5iZk@|%o?$UDD&AtBN00IMK{ni!8IL+mMgPTzOuRYIOuA?9x$v%R}hK%LhLEkzHS zdm|S~2A5C*&#Oz$p|pKOscx4f1w+L9eiP+4x9emtH9*QUQU8pp%A8`ikYatNu7I2w$#f%lAI)AANJttz-2 zD#`d8<1`xGf+WN6ujf|rHe+n57!Hzvva1;fckr1sEX*YmG(A77xpEkrlVZ98q`f4; zDB5fuNQSwLLT&9DXj3weJcZTxJXncGDPP0`3l2u9y!Ifw!M}a0P64gOL_L|tpC;ju z@tLa^!kNTB&6A1~-Nf2B=M-pi1%Y3s6S$GXPtK@1s31xnhbLD_(t0IaCG~!K6!UGc zBEETkVtt4SHQ#ZE2b22V{vmmN?x@FKyGlJ}r{Pi!sk)7qytCTFa%MjXLPsk=Rs1lT zS=kCZvY4rM+E+j&)F88XGUR0xmTxH4mSI>=3Zn#KEs$@t92TwJa!oKajK0iyJuiQS@vbqPMdJBbf}kp zFCOSRFMSiEn)%mr`ZFD!KM<~2xQ%gBKMQ$-xCvKO6x#&Morm^s_}8f#`8Xk&ribxp zcytzg#sx#Wrt;gPAzqs%Q5Y>gMM&HraG=o?mwvXclOHsdeARaRXdIA$cdDRqU-f&;Q}H~32%B%Loc3m?L=Rs%+650)uo z?)$nuO$X}7lERSk5wV4W2J-<_S3WUGqMryyPE*4M2HYvrk0!BU7S$)oO`X8BUo9DA zp7gYKx0;pE)3SW!55jA`kK4RHUR_dJ@4RogV!bseW`-%2kADr;=3J;bXJ)1TN+C)E^%8cAM=YC21Q48REVxqyFgLWu2;XRHbGOMAf0;75ey_X}k0qK@n98sf=S2^)V%Vyn4bAmR8H z@+18P%9kmxHiZjwL7>(Ck7#JB=m$d%O>XafGOE4B1|C;=5(n95z%=&|Pf5^YYYbk~ zOVqFIvnAE8h=7JKD1#?gB?A1xm92KC082h-^bUH^f>Nso4Wn4Ls`t?7hlTHujZ_!S z=8Rw$yy#EP9&_s~i4P}#Rqq>;c^YZOlkrnb=LqXZGFNg zM86(q=q77hq!;R(7yMqy`C_Z(M@)dAZB=DV(eMYSwmLa%+sFVxz$pL(yx{i5s?x_h`qoIjTTQM_~)n&UwtX$Al``I1l=vppBA6P4@}% z(BIWtx4Ko3O%mGG@CT{3cDe;?hNPS6BDsa*WOvm%w4Mu%bU!?Rr9X>CaON3*qnj47 zG&DsFTm_AKAgB}lFq~Wmtc>i1(;L>4sdOgMJg8kng8{$wZW0Wm`6Re!@q9un9^I6T zji!VgXc>bKb~izo32TWl+w*Ub$e?ARkrxH`fl&6Uvlt#70z zT{QLAWHRaMa$V)Jja2^fnM`^?; z!KB|x>C!7CpYT`{OImTeve*doJnBou#G4VI%gGo=>^Gxi8J1+`wx6201)u0P z&Y~NUVCUg7xGf2rw!2TzH+!Z7@`qlwTSX4>?gcoDKKDNbcU;GrNelza^>>30XhNc) zxITENF956n!%3x!Su&ti;8zY%pOgP{9tZEuhRWyoyg<8A6ELbXIC;2{NA)YM#PNP{ zWxo!mb=WlIHu>C=qZIbTh%5JZBdDg=}Y%kKXt>R!6 zM(ek?qzYYQP-pFpZC+c}eG{TIbWM{-MDb0OwU?f>kNO z2GXi9DW?PeTqtL!`2>9GnTig7U?b_QfE4wOw%$sN^^q!f0$37S$+vgN&;SRK2-w%p z>Xh{hCI>OSy%nr@Ii@7!FJO(qG+2z;x1R;zMDEfE-R8wHigUp&X(65jH#fR9E;#*_ z6k-nxR35PU8X9plPj`1H2VHiU3~erA z7;D@B3Qz)PLRjZ^3FWW@6Nh+p8?fBDPq8zS4!3?)my;XKJ?tCg;t^)hkhX;D=gNGw zU9|$*U|W9J9s+(7N##5NILEHH`lMT#`otfA+1cJHs83cSk`!?QPMQE3JZQ*R0Jz*` zQNxuqbRV*K`3B_F50cv@7B2faJ= zPyp2Cd<3!*;99~Lu-auT8Q$p6c`%*wUy~xrO>|0TwB09AOTP+}SQ# zhssH6J>(@dVizIX%J1VDYr>YzoHPLI)_vmyLWU(`-sfC#Y3 zOthunJbS^^r>wwR*Bgb2Cs(8p_b49RX%x+jdl~<652%fRaIL%41MBbNL2xaDND|Jb zL9tt@fbL$Wgf{^=45I}Vo^l%|x?MNPBrqEs4c$hh^ygQ;Izr}`^lY{OeXO3=Rzkv{H9Ej%$Z8^$Nb$ zH(WN{FhUa9*Sq2NNs+|87)$E z>!DwZNI%>5*?aA^*GeNnCF!z-EKVB>2Ti?uWLh!ajFE|)-o!w*bUU?ee*D_9NG@^?!r?W1s^0wa+h0izAHYrE zRRrg?UOX6AFgCuQTL z?S;OoIBCB&eKqT5{f(H`ngwvj$LNeHi(pV8sxrk`z#dEaz{-!TvT#BL``SMj6)*MB z;1Rh{+9Qe#7u8IO-elodZsZ2lAn+(zsZe2>D%F>_k=A*3pHX&{qcUU%=1~cFy@>io zNqC-+w~_s$Oq@p090hsbAA=p%!-Q`X*t+xPjIRN;mWZ&qQ+pR<#x$4&kQQJrK$#qH zdi~^~_IG~w-rEB%G%mEDdtl231!=L|jHo+P! zs0}CK#oyha$vv6=5T2jk{zG%6+5EwM!NBhyP+=ea&{$b1$OexErPAwSpW(_V6OPZM zC~Od2KbITXVs#IB3MiQdaA@a&ADIMoH=15zU=qXTEIdA+=ff|XshIab%7#LG;C(1J z(A08YAGuEF`g}hMC@P}D83aKXPZucUEL!cIwBua&c-MRn&C=>P)EIN6q$>1NP z2!LEbbp7pLY^kn)GnazIE}k@SmZUW@m_UQwNG8*S$^TIx(+Z~g-~MG1LJN3@3L^s* zoB4jXYy?NxY&;|Fe;3@%lPgk4dek6pG>UBc4$kWv+tJC7={^PH3B(`Z-A~W)L(6i4 zp_#@x+1NflfwGZ~S-JLeZ@AjabX=+(-LEPkcPSK&ozGqI$vFq48@w>IsSjNJAWFVW z9m|bHn$2f09ltM5nye~bUmtvia;`y<~clTNTCs-W7k^YFlQAYNB@sQ$F z9$Tdxf0cFGUZhIgldF<-{DGm3R--_wTYOT^li%kTB0vY;jVpLO#LVFQQK|D3pOcA8dwSSOO|ptTz1{lPq3)e`sU_ z?WCvq6R0;#V$+XY3^Ukq5b&@>|4Jzoae&~$a<7s*nGZko8nw2OH5zYK$!oHq9&SX1UdgC-Q>W3xp2 z_lO@ZX!P}TH#Wts9l_ZsSdZV>w=w)r@!x|h+S7Q#BvoB$uFzx|3~J7l2A8Un)qDL$ z@0MO<_(U1{o)%;#7<5dn`d*u1v%A1(r-#v%z7$`>^tg)3h?yzS*1c65>y@cC{NbC8 zZqZ!bRz*ki<(}nPHo=M8o}m&Tzc56l{=BalG#*7Xh}D{MH6{{lzfaD3p?1VNsZYZPhB$z@oOx(xQX>NmaId$C#035Wh8LrFuK zOCCz=8Kv6G9EwLl;ReeVPhZ4_(csDl_GU954F_%C^&}&!Jw^u8K@p}zO&EK zLiny`y|r2uP^ak+5NtMzSV;Y&cvfe72MPgUrrOp^ij4f2W)wa^pKEE4t5gA*daN*D&^7i~g6T-u;8gvuy=uVOg4V%a4V%vPx!Q zeKWZD+doeeY;{BRotnllMQ1N6AoxM=|Atq<;Cueeko!?IzT((l=(h34Yiwdf`1EZf zS)_W(VlytvQs69LW{m)>q)%5IIe!6#TiOuYjAOqX;m)WMUh5)}TR zy=J$_0~5O}20qTr#+I8?P!8K#F(Cd6*T)D)pz9dMp+Xmyk zbe14gk`JkH$Y(Kg>@vg4_H#HZ-**9ZD^utC!y}ukMG~NHHy)<+SgWDhyI62$VJzw0 zrf37OM96V4Z&xMr`$HsKD@A-)YeAeW6EBoReh=el z8m4fEiFwJ=!V1u|3$Y^9{g~SgC!j7ETZjsMgd}vvcghu>V38IbY){Bk2ZBOk{fQ*S znIH1fpGG6qqpWn6yA(PNUn{;mxgx>hQSP^eNaEAIs{vvZm=TjpDXQp7m_%m|3hQNl zca}(;JLjXzprCfp^1m1=+Odha3EV%^9>qq{DY{K;0D{xdvdUL7NSeO5key~#xR4%Y zeOoIgp`Uvy=ht`pGzq7AgAy=0_A10pD9*{WW%VE{B{LJ@$Pt z8nFusLvA`F_SD@>uB{6hU&j=dR$W5H#ptzUjQBA<;XXr#ws z^(A!FBTh5u1Z+=!x3AcrB62qh{}UOgX`&H4ktg>w!hwtoM3qNt!&&S z6iG9!d@lK&JxiDrh6jn=0!TEcaj-l=_sEA57#HA>f{;+=$Cne+qgQvy^vkLSepO_8 z(UlZpDH+;z8`~4x+2)Jo?jE34Pm?j8T$K>>2TZoQMINp_#u_os>tM%7ghQHqJH7>= z6k%to2}izGc{GeiBF5cUxho1eJKL#nk)_$ih=MV?W9Pdh9?@<4YBhOsRbs&(vWcs8 zK^p<^bRcI7sX9E|*3JsjJV2Gv?fA__6N>Uj^)hY59tMv`()y)ngXKjyf1@zy`Z@Sx!zI z<#420CtP8vsVq%ce0^r5AvZrJ{K|w95)IGM_brFZHrCg$^=W#P*=@I03hosZt~RkJ zPaH@A08Hd!mP#$V&4pR*6#b&(P!3-M$UM%p%dHVV?>OJr_K`8UMV=s>QL4nOgQ8*5 zJ39-e5er`N5G#Wa_Ju|;+%Z@9A)JJt%^RAWLP3O?%Mt5q!SFg5=!Z=7gf4O}3;*EV4T znN`M4a6ZEbycS;|o_D{qTF)AC#h{qV;&%U*#3KFDUj)=-D4(G=Um%IGYRHvipTaV+ zE30H~X#0`6AVS)&%@)fd#NVu_-7aJp_3HipJ|v?NgwuNOlGRW*HY#Y!9hW^7Qv1DK z2=oX<2zVFZ-i6qOL@Ngi@Vw$ zQ=6AvQI|&(x#EQm$w(z`_=K*S`Xn{n4+WwHGhcu$JB0&&h_EcT@5-63`ay8)$rZV> z{EesDYs~_Ou#Co7@0ro~4z?wAxUA-0@=&}p7YiB^Sh|pm4z9(9$yk1Xz-)tuv~#iz zt=*thGtDWj``O*JC%f~1O=Eh?7XIc??X^yUif4WZU>DQ9t9)uanLGRKQ9$o~4dt0@ zW1pTj8c5)PeVgeOmL$DV__elyR0_>gIfi0mrQvA+X-}@md+%>J)L!ot?J)$bp)Bgp zsQ|X05$T<1p<5cAOMvc&WLA%z67+(?ShU6sMQsW{*r`ub(*eG$wH#xNp6b}d3Cw1tXB0mGitA|6>wKl=E5RK*j>u&f4h1j zpF{4^H~43%rki_M9%b7LA}9f* zBN8meR|Fb6np`+kuKdU-s(ncG;Bpme!_zd9z z-+ciiMcdLNB(@Rh$iL7SihLOWStb%Qy0qhYKxntoGT+G^jh3*P<4BEHlx7NpGe$hP z*!M1j$(Mj7S~Zi`*K>I&hBvq@LO0N?Sh~QyH*DF#PQBF2ofr3YKS^{Pyph^xSV{we z*KzF`XvYg7q5ed7JB?0((%q)1qSa?{UG+WJKDwGFBrpOB_)_l=$0plpL}o z;*h{5BLk_UQ>xVo5)u+`f3VE7J8?F`3_$j9J}gBq=YTwo((K6LYZK z#K7Pw*VWaCTk1EK!EZ6pq!|rVsYU!t`U5`BWuHEJts*J0 z&C`t!$Gd%}8io6|L_4TF?iGG@tE7eD!{NwOVyx0#9S`&~gwbyxmJfg_NtH@49 zbZ9eC#@O3r#aw6U^m-k=UO)EP$o3`+ZA}F8_y;Dkot<{0AgbYH{+*cMFg+88KWxrowHbrw{ z$yajZ9&JdF(GhtLe8G(9Q)j6CemrxBUEO4L_E3hQ@xUte&@v}l> zPp-=A><^#pGWscik!CTw2R1(!*9T5sg8K6|lPFZ3c=hry)8dsjMW|(7JG_kk(!(^H zX9NNmp1U|C1n00xX6hm)%Tq$ya3vA!R>N1u=gCzG5r0Tyx7{ho_>!K4aEJce4XY0i zrW)F)W*7d9hSLdy%RO0yF>6Vjn!NjUwJ3He7SA7tVOihtUmZn2D64A0M1wAqRIEFwlG~ z)0E|rTsTlPj+jQskx(9{XJiN~12o~XqR~;e9MWO^#|!YZd`TK}xpZDzcOkJj3a#3%wzr|)kcG^NVsQRo*p+KNpbtJCT-)gBBDY|!Z@p2GcV{ixU6ozPcT|18M zZV4U8q;8l#P&VlYBZj86_grZp9f5uv36J4faF>4VAb}FBT?xKg4k<&e*vlFwk@V_1 z$csqw2eGomQvk7UU!R2ogDRHG+6Q;hP8iG*T3tiWQqFlat{dO)sp-23PYL-DLnpZr zdD-1zLfH_W%-1bP0rydFM5=?NysiGFHwo^5K@MZWc>7K5!@vIHKk;8Tj)Bme58lB! zVC>i}Mtr~QAcyZs9oe2-m0Qdo0LxNFL0IJiD4pWCFvyX3z{?g{Y!PbFSWvACHmjCz zf^&M+0Gr1j2D3l^Nr#TYh>f6bMA|myAFIw6JsFx>e9rE>%bqhx!Nwe z4pI_LPC?cUNlY)Wj!<=Uj!jJAO!k0nkvM^YH=Q#4&`bpXp%HFKn*pRd+7apsT9%0K z!ncaz*#umjrIYx}^hz(YbSC(Ljd(H$;b@_CDT%s|5yAjtE{Gm&5gf30aRU8@r^`Q? zYvMRytA%eLM&~gdGB^q&7H*m{w}UU4V%e0ww!Fw9%KSl{tDQyxHiES>gD`k^^D*$2 zRt$P(z>?Xq<^fRm-6eY#t38%Id2`u_W_L{c3;zOv+tIhbvL;9Ki_HUU4kHL`gyYY_ zH8l!0JAbTggcER?iD;tH_^xEobbO&D?M_wDlBPc_cD36qpo!cQKvfWI#=PG*NYrqK zhq$*dSFc(2dN`R|R2jKg>A?pkqp(Q1`Nql#eSuaPiqAId-lU+EDFl^g7+jT%j@4%X z=HCUIv{x$woA`rUSGz@0X7{=d`Q))EE(OuMW!MLylnaN`KwFQtBU$@-!t|frNA4LQ z5lv*tB2l2Ro1~DroWj8bW%PJLwYrf6A`REL>2{Op^|$|19SpaVT}O!0lpM9R*Pj8@ zzfk4r4-8#hExHr3%JaUy&^5noi}7%+d2y`3<4?|4^KE8-qD@$2Mtg$+%)ilyV&i?l zLM#{2#t@><2!P{g0_FFbH=ad-N4&mruYPC6*Tuk-E7EIvR5M;(FND$i$F+a`$Nz}r zwJn#im^@Xt$g5jtNdH-uD`4d5u83QaPbL@oiAQxsE*E@^W)!Auzu#huF^;aRm<>PI z-s9`e7u=xkGu3DRD*o^PA^2P}K$^bphEB68cY{YIuxgroSjWVAiUgXLxJla@bV6W@@h zfsBKX{gdru2Cg^J4L9&4WsWjbqtMe4-`_LNBI9Eeeqqi^j{rpU0B6z}*~*|hvW#Iv z-ym_zY};KR6o8YtUq}S=l%qbC%%#m-ewB+e7mm0)iVZ!g=L5g5rsFfJ>&ej`3j%h7wu$1>_?a zPHqAI^o_7dz4lw-Z2){hm8lzrg zu12V*$tA(!Mm%CIwH-37r*Z|B=h_;P zT@XDypTrkD^BoFLcpV3G%2Kz}u*FUTwt#$3H(zZ-e@KR|WGJ?tfpUBoKGtnj1#)ls zLr)u!r=G5dk2jUkXo7P@vqyQ#$)RQ!8V(#p>f_x>yIhpIdZW+5|M;&2*#G#i(SQFB ze>88}tg=yBBgAZhTE+1A^4#&ybb<9LsMz-FfBWlH!9I4)HKQViH-}O9?f*b`ykso3 zeE}}rb_%$>x+q7>AJ*DvuM`aTH{!2N4GMpJg7J(h_afcBC5>4*w7de7O5AN^8!|jB ze3Z|a?U4z&-p;^HUB)?&rm$x(Ls5MWjDtp*$gLL(X7kxXgHrrnr<|&yHbl8viXVE zw&}pG?bgf{$)~E396m2|Y?4meUTFcCzT9f5(?~5BmN&$Y0!%%-Ks=G!4$j}hDAI)p z)PG6kTsxU+z--paGRY|!Qft0e^}1`R@!_i{BaZ^jjdqblRPUI8>P@7lK!x?N1IyR za#6AR9Ff!SDtCBtMG}oi5oV&IfFNsS&Fwv;mx{cLOERE!TcCo+y`7U>CixU1QYpoN zPXW{tc79I6P}?^kd6+1dtZ!Ev&*+=K@4y>tRXK2faBrjCEoilR$GtuBz*x6~txbH4 zM+wfxfF-ViHF;Cg#5o=jqNMI66cV-m)=oZ2Kvo9s#cj?mhxY*^AhAdgpu3~o9SDv0*pn*|7VUb=qsnrFWcPH})1z@7?-=M0UT0f)@ua>p z3A42t3Ja7RNb|OJ)ys4#^=r!$gItQx2E`S3hNP~^6X-(e)?nT*!6&zHdzwfjSyhIc zn!%o*taRCAHr#N{K%HtlU9I;b#ZWz zt?&j|ekYke)*fO{ioH50a)77b{tH{aNDncq7HRFn6!khfmVPn;2*L2-J*BSIsv`(+ zV0IM5^tvPzs1R~m#0h}rS=BIA3gyr=oSyUL{p7s%k+1H?aVb2FBVO~<_1Kduavyn| zFRizWfLPQc!|*T^1hTA0^JJ27te3AdiR-1Zr_QXs&tk<=qx2}@1c>E2jXhB&(H1c0 zA>kq!n}xX)TvvP-1I6r53MqQ1T z=#t5;)-CjP2(>3TeUMG^#R}RHcj=@?Lmk#K;_(~~)7@Pu21*l=VqYZ&Pp-(T?{5av zEL6DT6+eCpLm_+B36Z6tF_G~W^Fh&)^m0$rq06B(B}O7S69u{V{9P2XRPLB*swNg6 zQaZ#|atH~3`gJ}Dt|&?4Ye+DbATC`|LWY37V5W9u`ECSK%g;2=o?MYY@;8!cHrET1 z%P#T3E*1@k#75Y0IUlA&mqlnjJp(?7;i<{@sZT!>jff01_G5;0=<45Ub1#+N??+JU z$tl-iNG);@T&eUeneRK#$oJoYq7n<8D$M5$E?fD-pSM1|)m!@a{B1?Iw))K4^PdK@ z+W80(puhXW%`BRY!aw}^PiGUp9@NmIy!gAD0JGES58?Uw?LRbEn#~{F7wl^JfeFUw zhi0n?MZMP`JKo)>Z!u405nD#h>~lU)pjlqpPWu+-b#x-JmNmK1F%G})VxAhIgtU62v7PR#Uz3uN^c>@sm#G{Y{~S~FcJ}zDyg+7HcmhCEYF~d$DZ4vT8ck*um(0F^v=O2S< zH2RD%_qknJ);v3Q&~Y#c|K~qZzMf5z|M^es{9qkG2@+fl$LLw`3y5(tMWiqPBJk}O z@D0Ihy#cMbR)&Sjmb=xOZ9nvEiott$DNO0w-Wy;WsMv%C(PAuD zObzI~X2j0}Zwm2bzc5EGotC=KKwe&`wDkvmwpP{(YLcVPU-0LG&wMoJs(`_VvOkn! z>opAO`X!hc_JEo7td`P2#4b!vIUD`wj+nz6F}Zj9{ig{OpcJe8HR9=*DC{T&x}b;jpmHq92rG$){}E4caMj5#uTEr3EgB? zPH?u&cgnJtjA#g#MePpo<_LhZRpg+D|a2Ec4q zO1_r!>G*;fcyd)v-XFf%>J}A&Y_tJMEl3n@QC}W`b4C;9h4B!ZqJNa@Cz~g!pCG3^ zV;b?G%HJLh036s}3NDfvtzgX7BXX$6T`%vO+Iz3FTEgECQdE)av++d8fvp|8T)*=VZWTx6v-gHd#*%NqpA z*jigDbMqKX32MY|UM6Vl%Mh-PiHBMPeR`VI-zlNQz+0zb9h3>$ZxGh^=ClSM%I1=_ zXY_U71#-96sxnb;`hzU-{3>A6kSW-O&Tf&EwfpewUG`uffJkB5WomM{? z9~?>BSTIXF>3!^}#-x&<_;|o3h5!`$ z0Jwpm_(Tyh{Pth4eBejF{T1B~A&I(ndgIH_#1o75fnx|ec2$hG*FGcRk~`DQm3G5d zzsQrTQmFXD8e6OD1^wc7&n@5g8H~G)lWk`BT3OUIq<4PK`;RAOrc&3_56zISgSh1+ z>l`#$09wiKHk+kue+gEdHWrCARvzcvO9oepu;;8!v4ANDPPL!oJoz#$Wz!?LxL!1R zSZF*M{^KT?Z+A{GBz9c7T+)BgDgX~Acf6qRZm*=%R+8+}_GIJ8%bwXdzJ5OUhXb}7 ztper~rC+KIsaFse@#&O{{W@aWm{|*N-!jCkJKF-O0H@e-?)3Ml^MGE-ag`iY;Mp)`B6q{%zbX~TM8%jeR%Lfh1n<1rC($% zwDltqdjs~=DJW`E<=KVdYc4W031UGsBfLx)5y?N==7q~bCw8%O@&EdUpCy%-3>;5d z17wU$Yv5*Z@wb1TCO3d8X4TG`^fmXFus8h`{Jm^CjIzd`8C#8U>lGc}3*8F-fW~$q z&b0aVcyoW-T!C*{FBZZJwghQzgeAfZEcXHL#j=Wn88V*3en2f(ceMLuf^IHs02}LE z)x?;K9QuhNp_9tJodgjXM&CUkmbm5X6p(zgT+nnp89_EzcydK5KacXn?bRY**CQ?F zoR7l(#Q}1pUvEY+Ug8{{%O{Ru$-b9#igCVxtF0Mj>TRH(JAh=MNT6%PbyRxy1;UaV zIHz@X9&LxDI0Ya6dIi4UML;VlGpjVA9Zm~c)be3RCb;F z!vQ<3BG6$ore$;j+BjNMps-%5?C}yWs#Eab^h)6@-2VhWvA0FK)oH(^aK)M;zRfQT6BdA z;pPLaY(IoGbLR#S`PHyUiBh-$qZu1yz@>@rF1@|-ttfd;tKM#h)v7=U9+h-E-DW|6 zs2@q(a>S1hMXQ`v=EFEhqlRD67GL^mr>hk|qnyl=lrt?zfy-Rnz=pt&KwyM+X+SLu zyR=O{sVjcr21wV6MZ$F<)=#J~PD8C>l*TK&-iYx$xguA9N9EnlYLS1EVMcWWy#*8~ z5RQ9?&~tMEkQ=YtTtDG;@{YSia+#kWg7K(4E0?MIi6gR`myZ~ zWbCXq3m8yL#~Jup9{8~atendTLZahjrtd5jk$K$0bVD?B4d$k+Ebo%t=*+xGuo*$l z?iWN6CB*&_`z>MuxteKk&QM;H)082#6Uhdgg__Z~f6)MvaskdRR$Jfi_ucZQ!fQ{i zKu%b#w>_%xcJOw6x+?6OnT?oC8i}S$;Kr%h45sKJIv!;{PP9=sbJv2E7RdNA+F-^k z?`K$Zwq+^PMOuqxMfUq3IZn7L_zlr-1${*FReOVfO(;X2lCk1RRhL+4Rspk6tawZ@ z>&*gs_AkR=X8Z^=YzzWhOs{ZA`jYYEa(9WahfC%X83hiJ%sX__v{FTsxGCtVz$?v1 z)iK~+63ivZiXPf^beBkbRm5GD92%G+jhH^HWR$f0G@3+7s|t;#qQv7oV|}e)DEbi5 z;vQmQ`f?~)cSBGGlWKr*%f$SWu*ikWegNW142%PGGe3r7$^8$8!7Y~vqxM0mNjN9t zNmSRKT#;ME-{_>dNc|AcW@_?0Iku6!I%$q`8Bj!|%CR2m=|K_*#3zt!Cu>M8W94@nT0-=T52caf2R zpnVuzm&_ciepviO(P|Y~{K_2u<}ckwvq1B~FwV{X(V-T=UC5+6yiRNv5dlhEOCPs# z{74spb~g|6K?5}9ZWfWMWf#HK; ziVA;&l5V32x6E97H`(!nZ-r;0dKfV=8Pl3=gzghW@6woICcYTexx0pjhk?57gy(LX z9>}Jckz-`t2fA@i*bU(a$tyF|`&jXKFSRT=M2?Y9e6Pecy9lPhv7`6H1dAyZJV?h_+P0IAbafDJq~;Z@bGneXQV|2q;v0fBGp z+ds>R>^Qy)B@HvZyCQphfDDM(|9GBqgiDjKII*Um;l4e&0{Dosh`;Gaw^P*TZiKAb zU_!{t9%CKO62ZL*KwxC^xq1{+Hv1A!uB^XdS1dxHnfzkZt4&0ESve+S)DQNL@^_9_ z2dQb<6C^M2tRa0cpbWW*Rrauy85xJ?iJNCnu1JdTH+aEIx}fB6<}f<>hs&hX%#`f% zkuZVO>9~E&eC|XRA@czG#UvP%y#+eYKt{{ctMmdk{OQ+rSBuy+JNw(eiewFujq4aS z2ma1z3&^EzU2~}sAV0gYMv6y0#4>Yt_y&s&GD4Uj=B76eMPm*#IAbK8%xk?nDFhq` z7=`Udlewe-QgnMWruM6UC}m{J>$B>|6xp6!kr&gWI&OEZNTl-M!xq~dOF{AB8cJO> zacg0mvFi!$*$wWEbzG~b=UT>MP}ToyV$=s1F+>cXe5Ok>dQ3<4NnTVgi0k8a4oa=!zEW6MV3Chyz>?fOTdGR4!0bdLEHn55G2Y4%7!oq zvzJT3i%i8bmtNczj>+I8j41qu34g$)3Iokm**BwI5Il+U2rU8E{cJdwQaeqApv=it zG6g~(pWlV$f~UFSXR_V`(yqc}-Gav-T)EmrNpdb1E5EWRKIZf}`cKZzc;6X4MW2Bnn3sqqrA$*AY-l#0Q7n8{%gB?f za_#s-GMS z$UcpyAea4F0yVtGaMe_b(7}ST!QFO|ncAF6*c>j+b;EQ;tFW$iv}%G93n&AX6t1 zf~a7XCwK8h#HkCUinZsY2ATT@I%zPy3QBgw&2?|3$$Di<6Mqn9qsT9N6ZYM-jyj@D z0gDsmy;`l$0ug38gfmgacxJVY+K?TPLyy7^? z8LJ!DG%T8I-OLQd&cz0{a&*MN8=8{-MWwFE5S%*xzVx!o^{g#V<{PaQo?MaR_o%7c z>=x9YD5=ik+5BeV2_+SRXe_qg{aITf@i=EZaC7;+01?~UF}f)iPNNG-ij??mYd}9K zcA#RghHRXyLw$~XnZPpEfebc-9qwXGnMXDZ>>`wmDo>UhL{(Q}-j_OwBj`Aqx+bvpYa;c|x z32#D5iYZ3287E3rjN9U~1nUnVwzpDHIahH*5kmztWj33OA+dD8=r!fBCYXwbl6H}d zhygU+h!SQyM9f$2Jj3rd!U??<&@EEoC9#y2C&LetP9+RKEa`ZJUyHUDHj>G>7F>uq zjvb<`oqj}o*EZ_&SD7>basRNA2G)AWIs$R@ghcp$!s@CE2VFX%e}TG^fWkScB+W6$ ze5BOVI|E^XUfRbA`xYS|1n12P1XhQdA|Sg`lF*#*y06ZdCs(Bi@`%)~7RaV>v;!mS z>;^Xy-jZ65_6UfH{%k%9cYBXmfmFB5h|?=42~o$X?TK;z$!-;^gmnRB=x|A-2jscur%<7Bl2h>ZGKYVg!SeOds7$5l% zHZl)kJd@xI!$MP9TjxQ1791~!I8qE_O@>4JEW34(m%l>sCHp4d<3_teYQAKt+iv-h zlW(o?cO=WUK;cwMp&*yIW^UA>Qt9d zj6{(3N_0eHihQs~+DX5YQb;c}MAlF_3gk!d67xY{!o);QCQSVLO_D!6w$bPo zM2?)cmYr9W^Qo>#wY1fZQsy(?XX~rXB+iVvRkf}E6jD1I(yMt+$;oB?; z<&K`tSeiXnj$l?I+JJGBe&u6>SvFP7_Sl9&R&eVLnsArUl*Dp-FF4O(yiT!2)lA8) zhW47b8fL9BHH<$9x6vu8U$^!*cTRq!VMV!oA0;Rd`g<5}hcDDm5~Cjo(oOvk>6P!% z|AO$@o6}-nCRbEI;zKe9yd2I|oa~1pB&%?#?S0`6!VHIRC*x}@E25GcQJocEg$hru zNCxpJP~2E6;wSe8LPV%o!FElzLBz*Atgq#l%{em0ijTrwm-kd5rvla#6|7skn?&Wx zX>!FJ$k1HsWU`EwZA)j4j<+pro?Mlf@Q2ejSp|8zc$Jqf;~d`O{Q%mbb-t76%p}iU z-*dmbz=~8pT5{J-KOD;WN+|*StRLHFE?1B$EOTU;;&K`$quPfnek`FCdjBOTB?i)B z8`l@Zq#BCDLe*OS7_o#%L4&;e_Nf11u|*l32P1s<-j2CSt|>+6ac_c4n+RH@JO^NZxlw-tWfHQoQ*J zr&V9nm(FTcs4pI+nVW0f0+pJ3JD7(Hddruay6@j6Q%cIxt~a~T&?yx2C7mX|m+Esr z2*+3CA9(wh_zWQ&Fp~#dU&DYtBdQ5(sdCfCj<+2Uo?MlL;|~%hLJEqaUTtF%2Pjv~ zlEGZs=|_{W&S23|xYa|Yv2t%MZ$bi6J-UaEbG6ZL1p!dFypK=k>^|t82ldBRT z{xIL>+G;_BAXP>byoWdi9D{biZUQHJ!0InnNf|h|oKK!&OaZk~`_e!evch7oDQ~9~ znUY9L*Be`kCs*a*{Xw>^l_CL&-OacCU5a6M-kXs3DMj9nZ`D=956Ey7D@fdAiLmI; z$MsoUXDB_&(3b*xg=l1V%Aw>VcGWLkeh$oUjlRnML70dFhR5#ge| zu!&)#5LG$djd)VmLEDYby+2~GC#Ok-KnzaDGmthx=F$3mgMAt>4(UO#|H(2aFE-wl z+zq6Gk{El}8ITt3Vk4C;YM^mbpg~rQ_xgM;CPzq+I8~JAcTcRhSBq zQ5GN&l-e_;mw*#*-yFrsz{z6rp^0o~9Y=7J3pZK&iO8xd(ya<-!esk=M_|0s2$<)& zpLh>%n$Rp)LHJT$Zb{R*D)HkFY;CU>wVn`I!wEzEnOQ|?`=b8j_Jygk_n6HABpAd% z!s&-M${3INB+#a%L!oj&Gyl)JmGlXN8Tmy@lwzLf<~L27qpqND9iU)i4g`q9x^ zskdt$1(w_EtE*2Vp(nk~K2lS`*LWOXqer7<9F>?4>dFN$`Ws+7s{ASc+04YmaUO@G z(Pv#U4`|PPeZYEXfT-P3kg#_QTi`R$$mS6J-Ya-YAm|a;{Di@uZo6}o-9@dAu?PNs ziEa+slg+3-xhf&#Gr_DEA{UO6s5#DpY_x=0J%Z8|U(X3CoAN51aJi`1IZeJgSOlOm9wVgBxxcTr4}ndm_klP7y+(5GWX8^d>tT)`I@dC4vlHZHL$VR~>B0 z_q*cxa?QlMAQZxQ zNaw0~#XP2EAtiuJ3w%`d$jxu*{g>DbyNv2lG8L@*QNSiH)+@m6ajC%HkfPHpk|}jn z;Q@vzqoq_pH#(l8(_m6ma&b99rvYqw6y6HKPFEM6ScU3evGI?8eXTcy_yHIJZ)b{u z{VXS?d-taHE*_g6(KS=h_iv;fz)!|s2E#-O17)hj#)==Z zaT^)s3XC~i==d9R;2~E4WTz~iM_Vuoz_)=--mCZ~PD}~=C?YiXljxGZ6b*_gB_W_U!gqV${?jCse)e_!kR43StTkIV zF|4Eb^0H(-((t5#MkG~11I$@QmOR+mqlD(gpvr09`7;^{y+T@21nLitI$ z-Xs%Rx2Tgo7PpZ!uP(;9f@#Jork2w1r~ zSl>e8GI*?HI`TAEBZ#UnS93w+kKYdKq@cptIM7Tjo1MD#hkKCKbY0I8FPX4%l^HoK zcJE+u0E&f|bCNjrVDIA)KT% zah{W2eMW`!LKuioY3F9U020-ff;qV`XirA7sQxk1D$}DoMti&}O~%xu+tOS=H{~l^ z+r96TvlGu{0-T5JKv<{xW9^9hBYpQdBpdH-adaiGj`e5YAun}K{&2@`tLT-}NU}}R7C=>xwR5R8W@17iRNQ0XfM$okqu;yX* zg7G?i5%lq8Dn^ZgkL7`kX5$$=i0{hFJh?J?*&m|WMW*EGV)0mv!uN7K)C&A2%IfLFs^=x;0gkP9wu96d~rw^PhbjF7Qkx1EmAd3XQ19KDLmnm4BDJ z1gYe^-b4OPx+e@iccXY#_AR_IBRq?>LD{GCxdkP$9{MI40F0j64DALcxnm5eKI z?Mbsb?McHQ0t^4GfE9!j5bC+XD7qsmwg-s|gCA=f@PLrSUC@M=FSuN)p|n_BVCr78 zQqdGO)yi7KQ8EBL$&-;-svuA1`qPF#s1?Wef(G?4p5ZGw$}&O^TG#ep&`>1POhPQ8 zY{Syh#iNwMRfNKfctX7O$SXVWEFG$w$$BnQU})fm>>7n%MelJf1D79Tnq4j>ePOO> zN_$jsI~))~kU62erKu|q0!cg4&AFeH5`Q2JJkbvA$(6ae`~kA7D@DxWJv@;io}RIn zt>DVvOW75+eH)AtyMw0xJ{LA7@(|Q|m(tLq3leq1Kzf$Jzy0%gfS~WYV9NGF?dRw- zdFxWPghTqMDHkMF&X`r*kI_t<7BsdKk-MOdoR0}%tAT;vQ_zrZs;hw$NYiv)m0L4P# zE`^(zvFT53AxV>x)iG@6@kgcnnAIq#Rf48HmSsbv1ffG1ZbTKqBA+1GqJTAVi`;qVzES&XlGRR+sN${0QrgEtYCQ zg=4mr1PrKa_$3UkM)UfT`eZK|P+FuLaV~a{71J-|Vq_0IB5_TFs)r1oFjYdz2IUY* zhpDL`d#H~1C72`eBYFm_ZXhJ(FOYWASGmcPD^o!D13%YV1qH;uhIBL4c`P(*ya$qn zT;MFtCYdNM-zb!R(p>Hf4{(hn3UX2H;7zSJrGfyN3Z`0B0O6R4MYtK_m$-`;*_&E_ z9EYVu!`drtPqv2kaKSXrBUC^ucvehK`=qwGythSOy7uAF-lA>AtDWBqTZ+GZy}NYWdDEN`sLsAcYJ-#yvDY|RFO>*V^dRC*>O#sFS%$7I%;i# zaBKg&^q5^wfGQ`5;(6Xq=^z}Ci6d5qSV4-esRKE=$rMS~L-~YCZ`0~}JLPOCZN46c+b788oz%?&bo=SW;wVz4{nB%EBE)iv;9j}g^*WB_&E0plQ) zKgwb7g^@TRaTNe4_Rb)}yqoLh=lYn0%%_+|mJEotAI(;sN7dj)XT3l* zcnkJJRzX<3p8VY_(J=ScKezQtlG@tzD1Qc-C$T<`5}_a2(TTFrUx> z%NbMS|Cc&S(p{DemaZTB0)nMm0nr>+WwiXEl8tWB(()clU&abnr@s>hvjmPqZxmgQ z1;AzWdW(?<>#QzJQt6!E02tu=JFxml=;V$xnIO<;gX}c81AF2hAPv+bh^=7drm?j8 zN&T3sMQAFCBm1Oe46Qu_hI*-j!yk&+SnCu_AH}RR*Uvp*V@?q0_>`TSv@Fy_X%8lN zI%r}Im68JHzKE1h;f7o%j;bVx2;4Az%~>DiI}SAzoXCeMI>48yG|HV(cjXx_o-cJq z{t&yydJ)aq$V4H_nwUHYmRbR#R>=#{yO%Z1uJ=5@XeUA_2H-!S z(j>JHaWR8Kuov8gBOscDK!9O%6wNV^m(5JT2Jy>cd0kH)5$zS8T$$p|AC}ideCz2M zaj(`N!MfMW2%l9W?G(DHE#*!&hqkZb!$2Rv!-OXO(tc+8BVYc!A5*}<(xI4>^k zBEy-reUgXWeQ7pA?`0U6N(z?+0^3qp^V>-eZcn2>HlJk=q?1%wrX42ZP$% zvq>;i%Ah%nXb*vmxOx;aw%}!U&F>QJfbEwJkX2t5Kc#SGl88TCvDxhu5$5|=Bp$}- z8*+*QmMGi+3`9WW4(JD3>U7++=eLC$j`z-B$b<}48ES@1fPh4$m87Ka`g~DiBS$>d z(l$SWqC`4@)L==!z{p9XXe^{_ACfs2RLOMX$*+NQQw6^Umu@_&@;28R1?k3Bs=8~e zc6HfgYKXN%Hh-qefPMnQK`^!2x@@RnGR;sta#;sT+gY5gZnOSNcOvy8Dinh1l2AzH z6E(6fe-T0C{(WWb?;^GE5)0pMQEW6ktx;WvWQeezH*%zC?T*(p^$Jn0W~>5YLAFH45JYfk25k9TZCoqn2tB; zU3+o`;-ppYcvRkPH9H0KjeWr-XjO55X-kuiJ?F-ZBm{_wM8Ju9AEuvJU*2OTi%>~_ z_qU+R1Z^z-EpAn~2Q70H2%an%7L8}5kMfX~PRHXr40-oyfZ`ARn|Su%rKI(w=R8Mj z9lTctxV!1jw)GoU0@zFbhujSII1 zM%=EZY@6n{l&(9$EKp}sF6HyWwVpAX{*I!+o?MYv+263C)h*i9{dw%BPdkZ>EQMQ~ zL>xN8ayM#0GW(|AdS`7jq96VxG4z`K|Ha2P(sEphTx1;vAZYS;7-o|xD2^ixAnpQc z$&vm8#vudmPujzJACDj$!>IoDSD14^B!Ncb2jOtQq#fzJ_^#C8lMv`U1LOX_;ObPs zJFcZnc#~@tob8m$^LrO+(s^R>7oNyaWap zw_0n3Cz#jrFoqjt;f&(&Mr2lZur{*Fa}EKgGyxH19dqNi(~H z?|2bU^qw;H6qbZ-y4pk|G^-uR=9X+#lh~O%mwQngE6*4%e^*lH$u2lvmCd50?r-AJ zZWlRV_oj0QC8jZz5h0~+n>Bq*K>`{#1 z;06(W@;z|hA8Q}rKh<_7A@ZW5Qru>a#*<&XJ-H&6zek1HcA;N9$kEM^g73%L+gS~| z3mY!M6(AAvrGP(&P}fWX@V*%2J60d&0%1F%Qlzt=!e-F;QJ*1#vxx@kq$Ir8c+O<; zWv=puKYXvV(kyuIw!#qtu*>RsPuAn9WbET&ymuWoY7#98$0KgJm=(R~BY$$K1PYfuh^nwR2XQ z)7+tVl2(SN0Kh5X0F!2`>eUh2{CGP25;%>|!GxeYO;jpU3?RRUu%B%~9q2#QYJ6^t zxD<>*fKuIZ zbPCGh3UOol2bSM$;fYyJ?$9+A7$qEw#0q*WJAwEQf?GbmMF@hM{aSLeI%3mE6Rk}d zV=@Zyw~)7;Ax+{di|J;8!h;TyvokiQCOqs0h?sx?D!>{gq!#7BI=Cp;E;QSI^fCym zb_JYTU7GO+8+O{=0;2miH233=jtPm+jVNx3d`lcOluIK9B65GJ-6V)4*_KZ{8X;oY zmy0<`ZDlz{a(wJVWR!RS1mX<8B_iv6$ih6E3&fKEUvpMG5Tqb?V5O9jLDcm%ifVRS z6+s<6>c)0DodWTUldX@@$bUx=x{-_G(51&3VHi!hCe?gG5v2(H-M7mg}6j^ zS*wB+zIB@ZAV=b&fVY%q5M*^ojjK#X2(~CSf{G6(Wk>(aCYSLfmq9=mz6*73u2(^t&0@eGg4u1f3OKTMuGW1@V&0{M1)6LL z;jt!@^FU#>`sTMd)rZj)txs)>#p`hdDD{?}l(J*+nHz(y2*nHyuC|zD-BPJ;VD*f} zpNw+Q0uCmONd#x~E)eJmhg3LsBUXEQqmdd~?Eb==(ln-LX}Nw^O!4HZTsZzv&TgZ~ zE7b6-R(IvU{pW{iB=bE0qH>b5eBfek;^`%aUkECk+a`64gy9a_k)%60&bz*ibo|8uvUBh_Y0Ai>2aI>c+B0- zdO@MM_?y31QFzor*j3>U!unjanr{&8qm5ipSaa|0`#E~@2S#gWM|UpW4KjH@Iph5e zSkIR@gVYvIf^ooicX52>+wt`vJmVtxoLmnA%8l=rbL^nL`V3;?_a*vn6$BVmJouD# zT`jUC*_p(50GrbL%qR>!-sXEe%+(VXSCANTf$=Hgm1Q?C2WA|H;|VomN{F;=t6ja; zyIS8L)KCXM2uFObm)({m-17E2=bmPZSqrpFXo}-Nm-CV*hMu!i|9vsERt0HCiy?nV zYInUz{ORZ|AnK7xJ_BsL)pRs)mB9TzXX?)et}&|hqwBDwT|R2xGB^v$&G9jpYhb%Sic5w1?hjsD*K0bb=!k%w-+yF5r6@ zYKOes4b*NtqjLDZSZGvb4%74pkghh{1xtAge>Gy@+?1ifF0w9`k?J(nmp0*H(C-m* z(b~PVqlnqZ;U!`rHiOHz9X)xAt5-?*18($ja6RXuOU4mn^*&|GoSJ1q)iUO(z2M56 z`;vji_kCCy0YB9nC!cHm2|X?xzS~lE7M+!L>ls6q?+c?=x+?tY+`FbfX69dX*QL`PDVhJI@<${t+aYGwmksAO z;;%|V`~u^9(fCR-y4oFI=sHiX%5CHiL|r9zo~8h1Uf>I={Z6E{3XR_0X7V;q^0;@< zAEtO#uO%g@Aif|9ke^_l!Y@l%z^qSfh-cDKCP-BOB^cHYFE!nR32y0y)Ozk~4Ds4};7Ti9Su(hy8#UHBK=oSzF7Vr0dbTOlEKv0NR zTtOr7id~F2nbMqG@{ z#lVv*lJko*>PHM;7X;%u=l|_b|rEUhy(j0YPZO9hNRw2lB zE~#kYN^^BGyW}6f5+P29ysRrM5P;;7e^#7W27W2#LG# zmFdS%sv#1qD1X62T-oF&vP`NbBHYZm8xUczgGZM@vdMHBT*bH3E1+454Xa1=5NCHl zE-IOZ+P<(5M5(J1R31sI?ScvZrtB?JKOJUhvv@YY32R%ln^cy|@*^00N6bGi=Y-o$ zjHnvKF@QGZrCbovgQT3o-q{ew7=u|fl9ZNC!QB|OmE0febbL+nd2&@k!6T5`DXLEn znvi_@Gx_0RzI*PL&b9VyhT+zNWcQL>=P6F2n~5aRmSRE0Fc*~DQF2W&9CU53!y3!q zServ_m!UFZZmeIqiazOMn&>ou*ccWcwPfeo@wRioldBRi{wS^8Zo#w{nT#X$u>}h?4A{Fsd8(z$ zGcFvP1L;W?GAfIc&kCPaZsBz;eS*^C9kFRXVf|`)BU#mp`&X>CaoV& zHrXe+<}JU~n{QARk^o5z;6W3>MxP=$S5--*u? zxjeg(N}Zg4zn%UN-CTGt2;0yAJNuq>Znp`Zj#Rg^-SQPop}U2ez*2kSu<# z>_yRS`WmwE=KZp}~yMijyF#s3rPoiYefmoyR@u!}X^42u&K5UlymwZy`MN_Od zenw;FGi1PM#8Cfz5)Q|7xK>r5e5gP=#2Os;Hm$m$=FXL|AD?j@u+6~Kuw;hu zWX%CMl38=Sst2sVvX)2O6$bUwz-up=;q9w3)L|eP?y!q!1V2j5{Q$wC`bUb-GkOY? z5+CAbPtIufk;jAz%J_|2$|$6G_@MHRz!IwowPU2Xg3n(xrDM*Ygd@13XY|rU)b6*x zvM~`%V!qcK4<;;Beu|KNmPAs?@bP5ULHJa{ti!^`Bi4$u%hTa=@OFQ@|D!G=X*_Ri zAD>8Rt-b(CtZI2X7H5=fv;bT)L$D>HV6NWD-C3u}T$Rx-QF0o6)di*xe~l!DVYmr+ zo9)$Q9HU@-qhB!bBYi*$B2Ve+8*0OjUpA!;>fDl%wdQN1s=HQ|rpKd-Z==~LmmA0Z9z%{cT*3u zToi$)5p1+W+MbfaAA_f8dNtjkw?Yf1|khl=ch%}Yy^_{ zESa)v8_QhI-EsFbU@V1)@xbA^bY3ZtE7f{NkNr@FM7f-#!B1a(vfy0ntIgK#J>Bo2 zn+^-gRY(~((q$&52dQL_#obTr1(#gMTYU~?>1BH3wm;;w(diUTZ~~(X2?{%iIX=n~ zYmM*k5cXkQ^UOMTMAb2{g{ZVaTHHCnuIJK|SN9U&*Z#v`_UAvT>m|YPD_5+x#Gz4> z#QgxGg^`5L`?W~U!$|`58E(;dEb7pf>XlJ#^G&MPpmIf2M4V8*1hCm-tc>_A_HMX! zAdl?L0c5ebVO@sW_?K`trSF=A?ApMpmShjhQnh@TZU2JCxn>`J3RU?@GM|zBu+b2gg~XMLM*b=rG#*fDXuw%F z{Ve0ylPeNM9+Ti|tH2X`i7EMrMeDwYl0BFnNvJoTelc28E?}4!Y2i~PFgFrO!(qQC z!Aw?P8dOV*{KXs+IkA=>n8%)6ky!9IKWeP47gWZ;H3*N|CRWUa;ATEr3VnMWAW@km zoJRvg?ccdqRpkmAz_`Bosgf;EGd0qg3F+LoCRYm2bHIK`CY|N9Sp(Xiu(4 zQ23h>(Jd7OMNtxPyBBY@tYKG={haHyH{-gg4X5tyxpdKcM~56vE~6O#XW=;1wG)Vp z%%oLzQ2aXV_`OxYd=gL*_21Nv=ko6AvHD4=)01L74=7ohcp5AcGp{~a%t{k~Q>SLL zNS=DhrgwW}n z34`_5-t5Z4z~7Lm*=!eZ>g8gB_4X{iKT;^Awym9tN6nH>G-C~Dvg?i;#~TciL~xVp z$R!j5N=X%_P*J#GtR2=3BN&eC@G|kx@HBPoGGBS^sbdxo{-#K5>J*FwHtn2}iwh&p zx{F0U@&zHXAlGh00kTw+_-qtj3#B`ZZ^{3ZCXQ6O!`@f;C7xWRfC@@nl0#ZRbGgfg z*x&-OpdmaEow!g%YH*6M6BQ|NxmBfrv_P2SM@5z6|6JuFP774sk_o2K^uu@}pqj7E zc*+9G-;Al*DKhj%Wf=oCxRiw#2@!M#4#FuOv#bT=aY8htolNcR_*{tH`D~QeU}K|Q zR&KRlhDdG=L%9g|6YvoFToltLa{!i)=A zHul;9?!=6^3g*cb){s|JQsYu|Cyul0XQ*UPu1J&QQL?$&T`SlDyC-v-@n3y6i3T*# zTS#FA77tnzaW@`Qf*am^n_@5VuyPj{SJdYqG~F!SL^V_IrY=G+HbI^(zDd8e6%6MT z!R#mVW69WLzujl7G!sWx>CZu|Q&hgurzMlk+A~H<<;DWpJMU45xw&4nP@D>$*!!N| zD@zu`bzVrl-?ucOA0vna#~G`^20_{Q=b=}__Vrmjh=_;d5CGrYobGj;d0Ou5Rr+eM zObybH&L~j;S4vT6c|F&k0hztfq4|R`TZKd24`*Tmn700eA@0xP>AfXGK3nJhwm(Vv z0c##>D4pW2{=S*%C!Cg)vjscAk((mc58sGG2`BYKB%Sxmi2`(2nhig)H1_1Gq+@?b zW~;SQ(0MMaNj)+7$k9kRN6I|Com2FQnC0anM3;1JGvReK=jRl_WWUCk?*LY2rO3{% zIt=Fp{b33PiCjzAh*^@*#OkdsmwB5lUpNI%u1dK0!#D|-f;#VB_KIYjh{Al%8$XTD zMO|g`F^kbC8%&pEjWlShyZd~>1Ia~=IPaI~A~3@b5sFd{5%Utl&Q z6u_k$mZ>T%87mz>%xO|frz#jm!ymSZLSF$M-s|_@_D>`OPJ}2_G=B|lBTZll71ae~ zA3@BLO1loanb;&Yfa2P`;rXWh%q|kX^dk}& z$qYhCPV}kUs9g=eopj4yIT6>q1HnU#xK8#@J1+(wh8>rW-x>FaipR zV^Bb$h~x9Hx!V6Mh**BSvS(!4QAHjUhUTSDK*CLIfgk?up@xl`P_$YrYUE zo?Mly#2?1l>NX21kHuB8B*U)RhZ~n|h3B|_X|%c@oQc`)+QNr1b)QXgH%e;%Zta06 zr{hbbqHbARh~$-&jouSU4!UF}uO^~w(ZQwZMZ~PT!jr2KDgMAqJX{Mdl0FpX^I=`o z<{uXuHJY~DqNY<8xMolLfh68kO5E-@b-X>`D5NL2chr5p9p5B*Wq;KaHzXq~RhhJM zJMep^CL@|5+>CN{243dA!lPN4@bHH{A`4e=IsFn(Sir&$9{h0GS)Bb|)+HP;@$`Uw z!!D3=a+kcF?Pn0KUIsDnC_>zBwhQQU^5-_{zaxC#3kB*;T7q_&MlEc&?+ZAq^OhG9 za;cgTQnu%o7=cc(z54!qhWvR{KW;Zy3+9t1=HgQB(&n1Gx!E4F zQ4_D^_;NM`F1prxxO9ILH}`d6;T@hXx7o7usn}9h&FdcyYocV_$BcyvP#PY)!x(dC zGjHDuE-v_QkQNMlDT$bt+g;)fe-8WQh2C&~m>_<-1xt$$(q@X}Zo@)}Mk>g1FRWW# zj&IzIkj*u1H#;J1`XYnJ4(mG!S0+bR@(q5@*7s8?+RvA^90XZpg*Wj2W$YqkekuJy$?(k5JSgFaCsF4111UO;)u}q zRDAK|iqz&F1%}&e&4O{>A%h(;*QRV#SZQrFxr-WM%bIlUHz!BQpnAm`5idQ z>nyJG1$4V={2Uw;KtZbx-4_9g^@xlEKrLyiUynG{?_$^`EL2#TNPr+XY%^VRRy*Ln zKQ0?1o*eqREA7mo@70up7k7?76cO>u0@Q$~cHIXS)``(fZC8Uc6|p`$7$Qa#aGyA8OdaYx(KG0o6L`ZKzU7 z(Jk8u6{LlneClm7TN!fS@yqWmy4OyKAxJbj$WHnAtb)&*1S+ReFgv9ueH#=#xvmdKHsWI=yrLLtN@ zCEA=!$sPwtU_8z=@9=S9UmSfVX!Gwvx*}++dtV$-3MUq7EE)y4 zt57dP%wgXcf;|I#?VXb8&?-c+_7ba=8PZa^xE>Jig~D*Dh2tvt`8%h;9c>s)8)@wvWvIMsO*47Nm6Os$BhI3_P%x z^b%S?FWa}lC^0TpE(T7>6bEnUg~rsq$7GXQwsx_>m+B1Y#J?(e*?1D|A`D(#v@2_} zTmBfDoz+%>+!C5RED@t_9YS(3Uvm6^4?+MEmiQn^xhPoPGj>tpc;QP!fY}fcugl?m z&$Ph)bK$~1Iy*SrL(!DWg=++6CUK}ne8g4^yHELINyMPta=`=P&J&u?ldE!x zctmm6inIi6>%#n)(r^GzjczgzN%Bh8NT;J8yH}P;oEgLi@1OQrrD|k2Pd;S0S}v!_ z*$WgDUH{r!i}%?dXxeQQ*%mNhnTp8*T7f|IXoMPobRMGuHo2pkEXXQd<4V?0<{za%d>x8bu|`5)HNDfyUC3vx7)&U-^Zg{b~(1BuO_Lq zkd3<@K9I>^fRge1@cL{R+?MPkx*cB-22ZZa72pq0MWwr7khc>KKJFi4Gb|s@^LZWW z7V^I)iqE&CVRMg!J&(u(mr^`!CuRf}*9A#=6sx=XjP=UzY9)DcRYJfYG}>(!*|%+e z*wWUPiQ-K7;Zt#Hpvjvi%2hYr?%n=U@EgFD#~O$;ura6x(otOOj$^-==3iT38hzI zTW2p2aO=KEBv^h|B;Y&>4R^cEf+0>emdS>Hs%&hvtUsHxY2xTeAv-z`L_+d+{ z9nStiS|NO*5wUb_v10L6u3%YR@io2U$yGUcf5_-6Y6?$d7JXjMQCRQKCLyQgECNTy zzK|>E+{?)(I8#z_!xbrFnQ1T1ZS?=u7X7A#XRku>^)+! z$q1^MHp9l(%<+V98ZdgLFOpmcu9BhSjrG4;8P>l?P2$y+A`dSUBG$N$TL*!(Y`0Vz zH)(UqUE!E!_he&}oPd3VOv!b=(q-6Y5c!aFa!gt|Os@~cjIJ8^$YZw?Lb8zrFgi$O zrK4B9Y5v-iUpcij*4OCAx;2lYzo@kqaGLwF?wb+;gD|H12u5?!&44cV%}_&=I^6$Y z6y*}gR(Qp0FFnq^P@4(NJ~;17lR@q1jV_siCE=1?aZjm z@s&WVbiCG&SXy;F#J&I$lbUgpHwsXGQr{~XDlI<}7=%h|wJM?FZ*a24yY_SuvN4GU zm%QRN>D#*-V}YJS+_*8nru{=lRb58al?$YCb-%ctzf<6<{w4YL|DXltuJl|6N$;2w!D^=%dIMNj-D_PdvaAG#NSMX2q~JWXp?ZOx4&O!!qywe=tOUR zT7-g!JRGsPekGm&t-9pmHWAe-<9+_r#H}}&v+K?dM8M_3W`r26lgWG}wTrPJTS#U| zKQdH-B~&tvtgd*Ooe(6eD^&>+f3uUdl_E8)k3a494(pryZ;yU7{!cBvQry68X(3WY zYpU#8z)mcfnt;-iZ&-1y;br^FSI|NFwmB zRa0XK?Owod(dob`i&1fL9vWD7%&2y@2a<4M?t5`jGSWOLClG0`uAHzaZ9R^JS_Su2 zHrcKBI4E^Hsdn&TXdKsGf_-&{q`30;h*Fq#`8S-b)f&^~Q1wDAyA<>$^NOGNs`WUM zYa-)02=|v@P6(<(-iReFtgn`fl!hne1b~NES58<&iob!->RQ1(Xq!b_h#&l;)?>2| z+%{XtS?v1dx}SHVt?q{uDwHMX-~Rc|WOJB4&hCJEC-ggNgbmGCX9PQXtYooEE(uH4 zw~eNsi4H-~eC>(OA_)9VbQ+Ddf_$-kjVun`2`=XGC9I5W;l|)LBvuT=rxB`IJ{dzN z%msyjboMnRh9N{9i7phEU;w3`j7OPXm%i_^h$N%cJ%gK&RcuSUPr{&cx^xVCAu9WL z2L}47asy=!xa|j+w9Mf%iIj>to@DSF#pwf# zz9*(Z3c-S(X@-L@1)KuD+_#lrnq))*_9DbSi7bgmGAf1?JIqGPm_)MvE9S|N)DwRh z43h~9U?;kx1nBC@S{Njkyob8JYU4(?GPSWs)!#u4(K-{c)$!DWe!E)r1_Zkg-7)|dW$Xt<4P?=Osf~^w!<-POaD2kyZ za9>HURCNl_zJ83zPHufIR;*fiRIS zJ@aQ*@VYUPQy*E3oJ~z|5>KafwsN#^PP;=K9fmWrK%iI0l2BJ)EORt8$mff{YM zmL(M;R_cOQ?ojmbVd}5(P71|vqe7)2>z`1iG;8Hu$XUTH0;a)a9Aec|GNzuh`c+PK zWKowi{9&~1mGy#s;B-$-SSFO=H@RYKxYZ=pTEXld%Q|}2!Mbb6mg{`Q$ zVz1|n!auF2(VmT*}Ra^|8A*uOYlM$X=m71>M56wifqk!Rcj35)6J!m-JNN`2% zzyaGu;+XPYX?Y+vx>stv>s%U+StJ3YW+)$w@?r5V2|IDl z-Pg8vxxcIZVUC?vqu{PS>TUM(dx}RxX6nq*-TiUv2Gk6lh6eZP4>%Wt8oe| zPuCAmcGph@qQoCQ*qOf^|b`4l_ zK)ZKHnKCDWv~i&Hu|X+wRqjr~O6B(t{-rtte<0s#qsRip`ssrnqwVF++IZ17 zdIi$gOJ*_eaLgDnUjZJ&p3%-hKYM^MEFu%zA1A_O?B>flHG zM@cXH!JKA7))AQ9YEwzjMkr;{ef?g2`{DwHTV%?U8xMwa4on#~-P!JWhp1F9Yc;p{F}26}tT zPi)@}RrY9OKzd(-PswjK6lYq+B_}GD8RSl)i@RVDG6dZXQB}}QC?^V->vTBrh^-X4 z(KERMF0 z`V+O%orP;Y>V}hqR^}s&(3EQpbst6ku+~PmXa%1Hac%ldG&>F$aZ4*&b%WN+scR>4D?U#QIL{b3 zjJGeu;U&0op~Ipzvf3(l*{epYXQ|4Tz}J?2>oh@`nH#H%Oi* zSLS;0hpsl(Rtq5c{{P3^yX`iTE?Jt_wV#5W7<5$un~~y4hs3UIP-jKxl%g^s6+oll zB0WeCk#r>wQYfzK9pYzH^dy(V0ZE?26HqTW`N>Av*#N;L#;XYIC1&=sr#AE3znI3Xg}KL4>+9BY ze56L?Rcl%=*f-^Qbf36FvYKXAq}xmDjZG%uW6vCLW~7xC%sKwEtTX1l7x2>60&l}HKazDE!z{$zSC5!Bv>KqHH0+F550N+^Pq-^^ z94x2|>t_~t-`{0&5N$12v0V@2(!#S#+EZh@xl-UjOm8R6qE3V=0$~KWlsE_Jn_cps zk`9F2Vn;{AlW8|8?A1-XXJ?j20pc2+=+kr9*~X9-%YZ}NY)50Agucyql%f68{pc@N z2ogm!54iIXX)yW=-s#3O$adev?V>YVnLOeTE3MVGa;lSAhrn$*8mDbUYdZQAFkFRe z57kHBS`K-zK0>>zvT*)+c+YHJKDpT>5wJCK!*dQE@s2K&a*^6X0R5uPaaqD4oueh7mge22qW!L6ufJj zzUc3b&6T0QdzAA9Da}!JPBCq&gIl)u=1>UI zhnQg2-w=HTq%0V7p5(PQ*4AHLUTa=6`U4@c_{~9=_M1RU`|%Vl;Zdnc$pH7l%`X=E ziA=UZDH6x6ckx8?EY%u{2Ei4QV^kZV@DbwZ@{W!tDTAs<)n+J#C7)p1R5ZR5xI*Np zg5gp3McP}huMBCg<_~*!I^HH+)E}gBaj3eEAM>?+tW!P zaW-n0EIbVo=2}P>7Ic?iNe~I!L&uib9%j$R9AhZ2o1_orJ&UG-?a2DNw}Weabq+3n zU}?RUCsnZ@TuTrd<3^W2lsdFG=hlu~I&u3Yt|fJ~1&3j5s*|IbmV%BDFe-uIVpW_Yi$AF>7Dc z$=imu$Ftk)d}hN@L#OP`U?I`~0_IC#-?O{I7N!ac2v{^CCtAOy%23KTsSy>Y`l{eP z+j>UF_-$q9*2;jA9%Xs!jcQKsA>XPqNxh&tgbFV-OT!Wqxii9rQq-_V=lr}kykqCp zL|ezwOuKx~-A1P<6Du=j%Xx6eZimSGdny>1O?aC*{jlVNShx**rdKoitI7zm&kHvPrTkHp8SewSHtXV;##y zK!6!Pf;pNn>sFs_ka5YK4Xx=l+upd4j6g2UPW;G#a;lF;^`!@q8Z^H{DdwmE@oF=k%Zz8iEq%UKaIrO zzVkjCq-`;I2^(sbqhW`PVqk|=Hf4C)l4)(dploPoU3 z3MT^LemBJUd~nTwg&@1}xM+@Qcss%xuWfW8R82X`A0fE0v7Qrg6h!gJEbl3(HJD#y z_qmaeY9`2`OMCa4w3$PKg7isEq!!{>fcrph`V;NG;ATpBgKrmmkwoU`CMa6fc#@4J z38k{JuYPR#gZ3K;l|P+?ko4``N1aKp+%z@Mrl||V_SgZiZ@UrC=_z^X6Coo_tg}y_ z`Y~IsNxq6NGba-k(GJp&D1PN(kjw(rncHB{Dw(2fH@EFMBfsKMi5^vd@ONWtE$6-@ zB&&DB_`V-@FzI$J2TyG)?0;IMTR!y^e~3paJi*ULsO#h{F*A-wbp@=OFBqL`F*fdN zhnaJRF>cuJf~xeyG%QwfRy?h)Zz8X{3MjZ@#2*GOW8tTtT`A}wbs=unF@;`ewdWz0 z%N*6L2wAZptPRC?Ek=*TpR4wviJn-=(|{gScI?2X%UVt(aFz@daZ=meHgN0-yT2W{LvZD zW-E3)7=ZplG(PTg{TUegH{l;#c}3b4f6zVdlX(E!(K!k{<}{!#rOi=f9ZV}V^v)!b zbGTUKhKm@LLJ%~v3Ps=y2MYkgZc5v|U^0Qf3FgnJ{^tQ3NF9Wr#3D*{o3`q;{U_Z=sWkBP=LPgbxsZPNOUK$H*g&EHaa0+MbU1D3Np;Mtzb@s?l;_Of!__ zHP_uKil9|C8h^|)dYQ`8qvGlM=6X))>BcSm$|ZwICD})08AXg96iXul{g?Uu!e(1# zgbI$hob4Iy$MLAjewywV%%%JLkbh&CH z_9i3ycxp6|MPQ?OD?=hHCCm8Dw6&G)mS8DV()HOvD!`*EHN8wGWW*dL8jLH1(yVQ~ zzDMq{({q**n!W{y_sike0VE<27Cdp*|9ZuxKU{NTy_PqEfKg}VzNsUj&6!FHz}FNz zZ$PuxNr=Al&m~)k*$M#+>NCtQ=oF%OY27#h77(pTmkR9@jLr+*mQ7!yhCkd#6vM8>ar0N(e^uY3jMt?*dp1pnvFt z>68pc|%2b3=&Y#VGX zrwKO9W@&K8F!`;-+)E#$E`taXMsOE(G^CWyW_TAEA_QDe?3}N85|tusQc*8aM^Dj--1Z8B<9>TfhXW;Isl!QV~b7*kVK>sQF zxq?CUoSOcdP^8*=V^wmIKh9;PQpss1c7TvsSV!ZKR(%wdF^OTsw2z}3k(3a-IrNy3 zr)#K=VcDE%NH3gvEE8nO;hA`Q`_L;J3HR{cYLi2vj3L%LV)^UXChW7QVhIf~Pzs)q z$di>ZpcYYeN(~e1sBE%CKOxe(9Le~{X{P=OX^V#hrrJe+kNs4hfnj$V_KUg0k$b## zFI;(50?;3fiIzgnsAT7C*S4838tIo^!F&NZ3i1>NY$J^;!ddI0P|F6Gz=uCfb2PJU zqDZo>>T44q)As7!!XiD*ugX*sc$H*l5GU3IIy6av8m3U2X=-*O>r{7z1wzLT)|j7l z1?^$)Mw2oCSm`RnI>{wW+${L(E#@sRvP`A_8m_4yDVQ9${P=7U^;@gpv!$_Y)gO*p zsWx(AGofJ|4P~FuBU{p<$?ESxD;`qm6LT+{@9Gsl64#qZ3a^6Zqs5dz z#I;hZ=jh9d1`IU6z4sp3NB6bT#y(q7a)RfawM~bGa7rXY;GZkbE<=RTM?t@zwN-3( z7}&oH$TLlxo<$}{)4Fu<+rJNsc`IZfOIS*5hyjwx}|LGu6;}-ts1m68&`3!m*S_J+5dBkQtX9P4Nw;i>( z7bWTt6cF&EN4u9zIkOTx@8fG0?tq*4_985X!8eKR~mU@ zRiMpa3;T)vg8MF!K%Y^}D9e?a{qcV*|bSn8ZnWY5)@idGW<35a6V-9 ziOA(YAcn$JFi4)W-|$VKL4748&8&v;2aQ+O*K<h&r+j|1p$KPG$^&r7 zW(V<*G0B3j$qs}=770g~(9~koV$nP~3PI93sXY^0-ucu@j!3RQVl)YEGFb|^iDWHa z1q%>+X?Dsbo596Gs^M)9gDbB{-tnlOTiMK8)@faq&uc%F_J$p~y6`_DG;u!7d?vF% zrtSPDrExheiXpNER!65Z@eKTA1rvCE+V@#J&eT|mYpbcP z{n%{ZClf%F(=KGlKrUveg#al2ZAQsGDHD7g8gO%Jdrd)0&tgrg!>m@BBW}hrZ2YvlA|LdRl7eseI z{^>9LS!#pij{o617z{U{-!C)nuld8#tJU1tlb+*#^ktcj!wZ~6#;Mqjz^qj^B@^P5 zw!v>Ej5!vATQ@My1l+rOW+_-M;(Hgtts^0I)g-csrtos_APVn}JlH!7`@uK`uJ>Y& zsC2@3$nwPdNrb60A)~YEAt`=9m|<%s8nLmRLkS)5 z7?%OE4;{avrZ{_uwe3G<4}DW_#+6s)Ao7PVSJ(5z$tfl#u(%T~0wK(S&EOZt6&xVR zW-0adT<|Ru7YEZ(dGCQX4MGYP&JmW0Y_`X0$yj*h8Xt%R&Kgm78Q>zmHAD-AE!KPO z9$K^i1-%rCm)2C=aIopVR+F5xaEmJZ6T}pbKbDuIFnJu>^$0-=FF;J&qE|V!jUvit z<*9fa-uC#97=8`PDVmgNe){B}iUcY~U10VgVS-QpjX;IE2|Wo6crg1KAiO z%gn)|nEs!EtA=O{Qh+c87g(5()#A^5Y2(#L&TRPzUOS93Lc(({i6IG80wEg#Oe`AFZFpi0zyhx;%rVE$sHu+ii#DC@M|@OE}JRzW;h zD7QzUg}2|F@EX5e(cmJhYO(XGlH9bDTeeGoVa z+5?ju(ypi>#8FK=lJChdxII#(OC`wE1SvySJBdjn8yKG}qI9)_5K?IDN=jBDB8uz!e#N zD-;CHR@mexJySKo+2TTZcIx=UBa;ayZKlo>A>i0UIRM>cl_-lFaXu-g)w#-4$9Zke{URsx!)jk+8e}^e$a*b zNsRKJ%GMR+^McXkNt&GKS_Nrxi!Of?!g{TdBRO!q*Fs(LIJge`<+(W2(}Tg;Nwf%d zW#dM|0b-XMi7t05Wt&wg@|6<^JOr_5QCYm`HVQY^lXSUVc|{VEzhPm$QOQY2i4vG2 zhrb8IE}ftkreR!6v6;%w8G&;bUV=1jF<`-Beuy?@f=)apaN;J7A$MirXOk(F|wIZn_hEwjVS2s>^eTs`GWu)>UJx{7z3d8PKhRg*|q{^C~u6GM{^%dxPEp+@1>*||% zH8i3U#&OD)&{;ZD%GNcGrV@=*3E#((Eml>-nl~;#04SO0(N?c2=LF5luJAg3rfTci zo^5uu_b1}%L2#Go!1KM_Vu`3yea0r#%QQy*MrVy`EvGSZauZ1hn|8f=`e%1h#l=D= zG~Z~;+&@l)pvPn15AM^m0x^>QJ$Dr>oX%=H>L5c^@NPUu%lu{ThQC#eMm290LwXHT zLBOPdNhNuBVOucq-gB@LP6s?t)?2Vs-^N&v{`BbN4C(~CTbVlpZFCe|lUG?=*HIlj zba>93{qqN?H~eg~Hwe*Qlxc`cu!FZsP9@5x^@9h0Mj@X)UvQRD9bZymATD}u!K`VIaW;z2bH&FzxvWb;MRSf+scDh&pDw9Mj zpez0m!}V%4XQe4waNFIxyg(+vtY+X1Q87t4jv8w?CLtr-ls`~#KJBgsunyk|JvkNM zu^T2I^y6#d!odLW1yFgO51O*2K{Oybc5zdb-WQ{!kujO3edxHyj3me6HDO+X6{1*| z0OTx;vrhF@dk0i}H?V2;hF?)2fcZqhfL5&(pw8bk-%vpg5rHZg5zm=mzR-j159M30 zZsnBU=g}ANy$=Yq@@s_U`cT$pkqWYF`Bzziw^WHY5rBi+#vG$7+jmDN;TCEz-4DMc zQUS~4lFU74Q`_mlT}^E@*8o9PgQy8&)oKs4ECjA0pg6FQ_pEz&-PZwx@!(D}rD!@~ zFkbKu*Zq)BwvZWL0rH7UHvZ7S4TM{t9#0?!52Rz_fyj9iQ}~g@w@Nef@#P;%ZJDiZ zI&7ZdJ7z$^a*~D5M?yurjra5vzi`~}4Ln;Gkl=PxG^z|xKYqf;9+1aW|9%mmW85cJ znMq6waU8hI4PQVZS6-E-!XHw&QQOFADq3_EdyfaENv{!>TR1Vg{$dnQyFKnK^}-C? zp^T6>I1{m~veCgk8D5@qggW2X*i)j)uV;T`rOQ|X9!|{1CQ3Oydp!Y zia*2=O`05vO>*BE9YwbZ72lBJWO{}8aTi0*#gSDeFy5)zDD5%U7Mj(h8QBj;y+Bi_ z<}eyfI~{2!hXWk;?sNdf@z?9HCl|&Ev+qI%NLy>Qil0>=yYi|WH~yf{%Eo$*dfG`W zosysDky`9G3m--8PW3x}>mi)IMz$8HB- z_BX8dBun^Lww%v_qc@A4Grk1ZiIh}u{neo>pMvWfTRF@KHY^K7)jZ$_O5)gbSq z(L~fqdkaC~&NxBKX^UuZ$2RQ-1EIs=JN`6Y}pDYCT6LMLuaY7G-5qd)A|H*7ZRQ;%R z*QTqf8}YvmpZn-kwya%0*I#RUi-!)MD;RX1#QDf$t02y2^>Wo8C$)+>PF{n~49`qQ zcH&!1BiJVz#UO!$vG#+Ka%gTnGJ}D{QYHGLjxt_i7mz!Z#bG31l~o%DLg7hNRLyKG zB%9LqDl;7torxYD6!U`Ro^JXgUvTABNm2f=5QFNDCu=_0y00#}9| zqaZZN)sYh^!< zjMXR}ln_08;15t4v6Ly4QoZv`U>%ZicaaW`!eUjl#jN151qcs zAC0{ES#*NH9T{L}2ymG8A$1Ab$Qoz&CJH*iO+oyyzVQq?>9En zmIEu#+M_wM$ zRjwfAt-j^03ffu$l^_=({t#xAw{!aKk2u-6f;la8%EzIyPco4(y^eWJlsS(e%4*y1 zAdWkXJ(VfCDCNBmz5*v7#aMjOMIAzmfRu}@*%ACm-sY09I*Q>hjWu5bZmzr{hml8- z^2XX`&S37SSE*0OU^N*)S}*tjsEf>pG%-TykcN)U0;VG46UCc;G2T!1@gJ4 zwwx=kN;r9B)K)ff(u>MpBBVL(L@@}BNjbDaz+UUM)l;Q1Wi5tG(Cp$VC2$u1C9Mo0 z(47RIrvoG1NsALEnEE!fISZ0y8g*aQ2UlK|o9_?HCBpLt2;r9#EWn;XYc(6#MZ%TC zN6KT1Q*Udc4|P{TdplvAjfJb}SNOPWH7P5EJ<>v8u)nyuFfLt+IF$brnS{2wYiui_ zGf7pVHajtficN8>xc*0D-4|VtE3Zlf`NMY`wY3~nPp46}%%4enolL$-Byj4=43El) z&UB|#v!yv@MuP$^ZAHVWqU(3`f8Pe}_!?nfnkuNhoIe<UY$51?&qZsy(Yj4I7cfs%qE=}wji-;Tn07~N?25uBn+Pw?BBe-9mVh% zGQ}tiiO&M;)!o{VrzdB5IWQFqjrEGJ%7QDe%DLbVc-^Sw(dA~ihE)qc4_Ov5tex@S zV!F36&+i`=nMi<1tMbBe`p?-YLJ8?Op3=r1OnXzXFr4QpUd37nZiDf?=0fpETHiz+ z=H3zeqDOZ+MKC)F?+S)b&DTM)URxDrOU)nRy0KNwfyKVtjyn&n<0jHqPO~&w7;^l| zvGpdp)wd$6{MmG_Tt%9=CDZF#Py1aadUjE?eUX%$I)6J*(O zr9t!G*+yO*Y`f(%8VyydnK0Q8t`#P`(@)q#OvfN&r(N`}hCMo-M9Bek{UcJ^43Pg^ zItU^1GCVpB)fE+Nb}>BkgmiG_RXLJ80;!EWS=-r>)a-Y|YrrrSsEgA+JB(Tn;1*87 zKcj|5)1vh@GRuS?e43Og%2L98Yr*Me;WW~=NDzO~ijmq~!Df7Qs`wHaPU`?Lov3C$ z_W1US^#S{`V9wd}R@ZE}ZwTP8Mj}5L?j3&g7Xx5cqCcdF33z3eF-;aG-Jb=y6$GohoMZ)#cQ^dfwAZI&(}|2=63f7#-KMtW(f*EY^>!Z|{U5&j-~NsT#_`|38G@eu-QT{MeWc8u znRa9U_V3?}``~7maTikD-~Ih}KmO@&{$~F7oSO_RfBpCFuj1WBlfp-r(~sK+qRu~j zm;4pq@vq{ahP&balH}~#W69a-`g|vR?H^5%_RVeY_QyZ{G}VlG$UnWb_vANme&YfEZQpk$KmMiD`SDL;_Aym3t$1dXmvO%<)twRcDy%$^ zCKG3-;TT>9K2034tYL;)%?w$=no!}-oO7IQ&YvNfaneUfy z|JMFFbH&}|e}3{NI!g8R=g9}>kH3Ha%CGvEtd$#^^+vV4`BJ*Y%l}r>Ev|0P!Na8> zQq8s>!3iWnT8a8N-GZf^o%i<2BViPo>v0ItrXaH2odnr1j; zY8qSz>@uWttAlvj?-UJ%N}+0g`PZ`?WR=&g=I5dCLZgV~f2)(QUQnf73Wd|wL6iIZ zzKZ?wU7`^s*H%Y{a7U_5GN+?a%u^3<3psmAHLJQ6B~2-dOb|Yvjx_KTttJ~9h}c7% ziIIrF{{}gNU^s{$=p3%i{q4r4c-qH2@1*7O=Nvf-FIEcjgbzBdXDqn3Xc%lDm3{+j;B z^ZWin(}(4MtKIi{WsXKK-S<75{qDtk>(ZV%v#f=~14}JqLCn1inp4vvA%jrFJJD)B ze*ts|Nq8S_9_x4-gC30H>8)+7+({Brf&mP>t>DkZ&jmYJzyOFva1necJ9L9E(IEm- z*Icm&4qK1^I)`0CD=r<5ffJ1KSM*dWw(Ye1YhDhiSKfA-htvyAIF|pdE~M)7Zf_~1 znh}ae^N2BNv03Y{Vi%lsr}uEnLyfa`=O*{wz9Gp{l^;Q;MHntWOk zhbf+Dk6_fHDi!zPtAHO7BN9?LC5gfXybt7EH08RRHci3Lp;lvyBoP za5|dIRWN73$T*9!1#t6}$z?x? z`c}SboarRg7&$nYI$bna8n9=gOh_|z2r;JeBL$`U;V{Octj(<(n&f>pC`_at%L5U=X=C_G8*Su}YjjwH zJ8cr2#;B^dh5?HTs*a0*ua%8LGm7P3#`2xtNM;nTUaVAdXB34Y{Edb6+NC@HK1Re- zdjo8?#aez$9flLG1)!|sRAhGxs69SJ0I3T^o@|6b?|qok?*gP~wuv~1D}sp@@%tpc zSTrg&pMk^9u>S9QlgaJ)hwr~9%5UC`Z-o+sqc`n%@I8niM4-|4i1=3U`1yWoW3yZ< zufe~WA>;n=(9K#-b+89>RvsrD{0Wn8X5#FYS2g=fbax*?;HOMJ`iS5nb;^K-?t#y? zqp>;PAc;?&PL~fKhQokmBA(aeM}9`x5>l1wf$$4u2YiWl(TI{DzN1V<>NR{QeGFK) zo<7_ZZC4VSX=xqc)<0lW z9jP#AN7Cj15QWJim1j>4BDF%bR%q*R`6*kf6}a-MM2$Znw6fO7fi{vhv&5-lvwYW9%$JrC)u63d z8rO-QVkAuGF1@JyooO;uPhR7@MTDp!O>&H&5fF1CL@zqjlwBfFbN-;&0A%g+1|@~Y zcOqLk4*(ilXXjv_D;iLZLTtR>3#i8H?hY)}H~tXT3K2CUQZ1#S#hD9Vr!!v{SE(Wk z6~}^J{G+IeETOm(v&WX1LoyBt8}*|_?|d=IgWtRJuPk}6aOeGjq?O9noZG#0=TD{* zy=2(WXhlk>8tW?pc*WM1-TNYWz%QatunmG42PEoDdX`YYR#PvjOki9Fv6e*`!X;Uw zWg^VRue89>AIcxZ2x&#L-ZbYlodhQBjpW4&o<&=}aMSA6s^F$CCB?k_v{-oy?3_az z!c3+|UvrCtXYd0`P7^TlE+Q$5$-?af5CKf@AOk?fXjrA(0P>&0hobqYR`rFTa%C4Mua2X=l(O?v`KY$Ath5v-NQcqC z=%`8>>l9AsU^`*uq@T?)!bsIiQSU-cyt77H-4WRUd7cz2L+#rpcy?|z$#Sr5`OF9g z({)iay;k>ysdD92i3NX{YNcMu8POzDyre<_7=f|MNDroBBrKhXI2vdf9`1-zhl0lG zhp|E*_0it3w^23?Cy!vgw-MM}8XZDX0dmwc$w1c0IysQl#Y$=W4P`JVV8PR+@eG{x zo2bvV#_LKG&dcDJ(oudYYt-v=9iOGFam>?R26)D&)U-5`vt%xboQM>*p}DI*p1@_y zCK4KFB8gH2yuIrrzTQ!2Oro3cA&_cEH|VS0_$eyGedVK8$Y_m?pLq*gR?6^vuFsx* z!7wT2c)9$WTLNBcuk3ia5GMY}tQ7?Ip6-)R!;5{5dy_>qB@DAm91Ds`2=ch+((kl^ zdFQM}6I)I>29Ht+?0CAKK8CuzKqh*y6elPQo>PD@8lmjLrG%Q7?TQ|n&44iN=thOg zkg}GPa`}eS3nDh3x!^40j{z*lc*N!luW;?tqT`*~mM{31E3ZgV`%*H@%MXuho*D~F zvrB2`f>LsHG^xREOQPr4l$yekgI-L7894?c2>|pa)2AasWfEbH6HB%?Szs-RB!^#G3pHPb}Nfqc`zkTwE1aB+|!!Ko|8PI1ew{8~> z5lMz~&W4nx_`sLi>A0nhUIvKEeUrlc9Zvp16py;(A=V19C{>d+;NW~44x@I+k#;uh z{EvSQKNCIELD2wtj!NCjFcPXBaoer+oDon{RjBP;4JVL#K;b6V<~ZYCX{yuG;PT2! zdmlMb9?nVZ2t}MNfF0_7%EPi-1x-9d_?0j1g5ujPWb85=eMPgF+eHm)P7g@DjPy#; zd$jHglVwYERSp4vWY+3h9;j~*v2M5?MOe_}{IL%@cA1n-y-HFB zGQp@J8wQ&2k@&l|iUsq)`6C+dOhjjvE_l!UTt9|4K(yA`XkgeO{~KNlcOjVvGJdAW zNQQ=2wAq{*Q~DnS*T>Tvb_IJ8eh>s#(by{H(D{1@#49^=F0?@YaM)_4o|Aa&97sTa z)vbuVBHxMM@; zn;>25ksT58Fs9pX2Eh(_fya;5A#d4OK=J6lr9CeiTwA`t+xph3&|hClPI>uBvVN#q z$*VpRb9;_2Ml)_0c{HR&BC5jFCL)F-Xo80Op*BcK87WUj0cfgsw-D~0ngu9j83iAl z-_ln)Y5avXb)a4gHniE({FJV$Q{xtlm`24H#LJae zsxKh7QC$@vx8{$jTdg*7fZU6)E2qT7f6?WPV-yNFf=u{_QpnkSE^cSQjdEh_nZRIl zL?LoE4mw1{_Ovqsm21i~Sh(69<3>Ssyo&^ZEITjRuwWPteV0EwMJLIPx-UX5S9U4m z)mz+}KM=QCUC#k=cc=f~e>1|%tk-qsoO9a-$8y<)7*lrid+=S(7JK15f~fxp!d>Fy zp2`j2L(u8n6BJ?>Fe*G7O~&OtSl`szl)H0$%A>?l1xIIFR+NmgV zTTp6y`S4!C*=ADPuOCGIh`Uut7dbeaM#*z=cXj$})3y{VmN%87iEE*Y}(~hgh^l?^rQ4?caO8Q#I||&Zocq zQ935L%5scK2WY^7io{r7zn=dp@i1w&EefN72;6Y`zfh?7!$GU{jhwoYp$qgeiKy|% z{ApFcdkTJT=vw*&Uxw}8F#44(iy2dYSuE?(aUnNDg@)ADH2W11U%`;zBJ`@m8Eaaj zmxK~oFjxxeYyVELytcmf920pdf#l_XtZQ2vc^pV11PmwTCX0Jv>1ULQ;F_^{NiK2e zB_71x5lg=hrG+BnQF38~V!C^Z2_SG~5rTr+fUnWY(wd(2Ajbi8OTIB)q_j&vlZvjy zT_XhDkGxgH#N?nC>Q_*}E~10G{M%frBWBv#uU|)aglZd&T>f=`5;-&GrJRD*j~lk6 z?JQvYllwd;kXl)`n2ym()BXPI&==S1zG$voc~$TXkML?^E2pl!)XsNLi6GHnM1t=V zlYTpkz*8NCG1X;sN@~Z^CrR%u#m}SUGVGEixT>-3FMDqA1}TlohsZHsFr^R_);NVE ztE%aobh_Y7-_A5|&;pISG$>6+&F8=4Yd5;S)~6Zo!k(6yv@}_Z5t!E3eF3aoc}2qO zr4*BwbE`Y4*6TTC&3-%z+k#IC1!KCEQ%VVxHi>C7DNI%V7szj9BN%aXF%-_!LmDTl zA$D%PV{2ie5fjLBhKs!ChD(_cJkbW3onNhp6?AA!9BuT?RyR9tz{wkG?_SZGczx3s z*22lCPuYCV+sk=EohgTai<41vWt`112o&q z>1Z&$fpT;mPPE#dZp34$yFJ3aNJbE;&F+3QB3~8FFI&Fg*Y&McfnVz$;n$5k9xY6Q zNj!PD4SA4BCNqkf8lxG?F^i*2hG4SSpC%=)wL?PQ$r1eN+9SBt1#kRD#TTf`l~*M< zcm!29@^oHj_cX0n7fC{{Oi7eXOu(d-bVxc$GEwjow_dk9>0~swRjcgUnJ1RHUQpa- zqvng}%9U3oIy|DK8=E=%#WN@NeN24}C7BVHn(a1ZA~phzw#n4IIqWZfj6$cUJzLUZu9knVDjr<^8v%0K`yI9S- zQGW(_`c1nTTzOT3$0KUHvEZ@1G~&gD4p`hMO5(hQz$@d@F6)5Y_^c*qvb`Z$@PzyH zh|csj?z&BjQ*DL51GguNPNpnp#2u06Z}?%>1cMPnUh0=yYLs*!bMi2zelERdEJ(!a zV;67c-vJ9>7nOJ1DCRe`+zTw>EL?e2g2*GDyIIZ2K)cbWPf346g~1Y!NY3hlGKq$D z(Hk|%ng>hclTFZG=~R75rvnQK95MULQ3vmc=s&V5(^O11hqQTI$Z?{FsBA0LmU&Nx#cU)Uf` zln)*O`X}s2XedWht*jttgHeZL$(r1y-O<)V{|>f+4mwe49dZ7U)#R_VHHD+blOG0A zllo!2;yB1fjmNfTvymeJN4ct%FGmB|W{nR%nnpqFi82_4e)i1Kjc$r}kH`J5ggxQOAA3D4e6}EvtaH=%`@V$_dG7S2UD7*<=t(D`1o1Ldj#l zw6&3wft*J#DhFa~n1YV0c8pT(c+rX=w1%m3b{Pbz8gan=MOe=4mS(b#QC2>x@7BrH zQP=F6YSutHDy!rPzWwI?n|Cxu(dU50_jfGRqn;w~xOc&H^0jExcrw`_YEqMpS1)lq zHb@BSK22ylf|JCAnN$$-&$dfYX0dcFgyjJ5=4#=!{T{Ha&*KB8S=%8=^{g6 z%p{4A;lpgj3;n-&iDKT7$Qnq>mS|A+I3zHNA#dC=Po5Zd`IjlBxGCd!L|{fgj* zPhJEP8tgGY00~w$q0E@FFUD7{Dv^*f@;Z=Ja04zHJf1u+2%Z)2yl}ze(et9V*2t+> zTh-muW((f~Zlj6l>)~%rs8rjfozv5V@?=fV!w`~@dgyc~aKV`Nd6uI>2Xg-9-!-V@?G?+ZuXb?=PjmE22u^w$SYLz@A zCs(g1f{c0Vm`FOt1mEc}g3>k@t1SoVnyC#YcbEY2lbOUsYs18lN+pR5eWPTRR4@{K zV};$oIgl7S?X&4a5;{{0H(ck64Ff9AvDf&9(IU~Z`T91;N$T;3>((lH+k%PAoAm#* zNo<5B+aMmIMr*P22xt`yMz)&DIsCQVwDk4q8*^BY4Qn7kz1C}pHiL>;=yCC*`7NNtT52yO+OP0bQ1_K zRubd{ZU4Y+vBV&ybef!Voh!#w83m6FPX-zU&xz(Du35=DFXSaQUPl}|<<$3&Z`MzyMX3pc=gM>7Wr;U;+Bt(Q`j$jurG94)Y7 zg+KHaAg)!a$H8s+oDUFFMT=1s+P$oT5ko6r#gW92N9>g#dK!DRdNZjeE*Kxfi9GK; z(6nHvfYhgE%Gh|0Yzx=M#txDNTY+ayq|BJNOOu~W)ZhgM7rzUZLZp0*Sk-Yeo5QJO zi(PKIPI#8IzD_ZT(IIcq?P2bFZ+} zH*^Nh8ubaHn)V4PYiy;=DLGQ(ItFDXA7z5bv9R$_=*wMWG7P$!@a?3MpDmQT+C8Fh z+$PB8QWxyD#n(X}DbNFUjo?R}L+8Y{p|3^zZ#Np-z#h`iHlkg5MFs#BkHBtiBWFAH z3>^ec8C5jN0N{Pti`tYxOo}soj&q$$oUq5`?v(er#VDXJY>lPPL79NM9Ajlny0NN9 z$O=}*J=8MDHr7Dfw}5Yk)sc;2XLGF27mOTFHXKAwYQypBIGjftj#_OqZzSNdkUOGW zy__xqxK#5Y3~tbQOBX>1mGphS$oCoJ5R%WF;k59dbUw{i-VOM-Qu=t>hQTdC;<#{b z@np_Hgrw#iuOA`)wh3$XT8@z9PJ9uX4AV|5Vj5B#oIY*|By;*0EQg^S=|Gw1U^{8R zGD(C^Th^B6@vU|Uov#nJE@N0z8X>WkOe;r~glEgBD1MJ!hNmKU6^uAs0prn#qgLO{ zVcH%C58?q5pTo~U&!uS}pq3J)AF2k^+W?6<&-2H+Tg(t5J&WjcB%1Ca! zM{F1U%BS07$5bpw!#mP9*astG+d^k0hh~HOmbaUMt3^-Rt_HJf6!Td{!%y1knns4o zIkLd}s4a;D0Kak2=IVSpC$#w|ZCBlMlNE5vNaDmJc-qKoWfy6$ z>5XRWhqCB^jTqOSIJ8|lT6d~779_Va2TAiSGhmaFzGF?8ilRBBJPG)pF2>b&a7LQ_ z@CyM*HFFsw9ts4KN+z{fUw?=TpBY?CozpJB_0g;09kDI>pkv8>{{{6OZQDw<(DxnhuER=^b_$uJ&q z)%Cpn&$%P0s#r1R`=J%4py~VIVI6h=o!}!SZy#hNp}~QlW-`p2?QBWL&HrOrixQDk zvQm8{T$+?G-NFaXx*x?h!eKp8YPJ-JAUqJ5eImFg*Qf`5wvmH!J z+v5(B{HYrg1E6@ETdm?f>@Q#HEMELME{YI%frOq>M5lE5UT)n}TJ^U&H4jd6qq*pqqIp== zo2({ejq)Ter@Rdw?7`?%u+n999qCbN|h( zyepUcr|S$x_NK)5(L*ia$qIv1kXm89ek$+?i*9b^)DkmQ66dWq3fbQ$6)Y@6(1K(Q zFMvH3c=U#jIAJZB&k7x3T-UeiB$GrAgP1+Mq(DtpuhNC8tLZ}UfXg$frLXzu*0UtV zwrH{fCDx5};a3~dS}8`c2JZ#XI`$qA)G|@f*7Ul}fm7PU zfM+^SQ*#yk-!idk>L?(-P*`m%OK%E^B9vXoKN*mKb~eSKD7ZXV>Z~R3od`B!5yRBr+`(ek0>m^ zoubc1FMESr^*ErV;lKiMEuGzsKXZ!R#CJtw!jnM;F_9W%y#8tA5mnvFqogkBklrdRidL7t%qzI{D_~yX(t$_w ziuzhTXUO&+|MY(x$C%z}Lvk3GC^g$ zj+xi6HlssgBrFWs=d|MKO!^c)h;vwEkpcGzPk89em*Vx~!{4HRy|R%* zHpbGCWPhVN(UH-QhPUY6Szn?GB0vFrqADWG46?0;bhB#jS3Z)cFg~J}YG>Oz7xBi3 z5vw&NW(1N!&B?pw2$@n)*~E(td@mR>z8oz!)>gsMA~|jR0it#0ayeuE@XG{Yi17~u zM+$f;mALqkWCY1Yx7I@BB19XHZ=BQ!b%cqHWRlBSKfD=75W}KQjH(*L6rw`BM4P=R zlYom|#J$lq<3EsdXdvj8&wi4FDPO2^6=w!HDBYvzWkn%|bi;b7@cB<*-m-4yYHV zU+Ro>y?%5}p3k$eykoJF}uzPvDqwbTpa_0KVX$Zow}&*2`;CjBxz$rOWVN=-3dKbrg@yLIyO(??wE=;Pjn zm?!D^Cg|TeOzNa#chkIhq;`L8^W7w$I zbM8}S_7rikZO{WznLfZ<(bcv=lPRK$2Ad}<4T5bItTZgx{PkndwQ^ERi<0x4#ebtf zu*%C(Gza{wR9>7WEij#QYi&q9XYXf1?JDX7>81t~rcUaKVOD~txMD;jEWqGH%nY+* zn6`s!YM!FO;>j$7U|9vT3=0;24HcO$Co>&Fst0#L)CBO)%?C{Pe$pGoQ!tmgjRu{1 zQr-aF$x060C0hITqr&ApjBZ)=(&U+S^XOXAv9#<41v(j7DTA`{!`XE~7CI!e$i)ciyq@C_Y`U)Ojs=H zWYe$7T~}C0egGj8A|e=ynjrb$B7=j?=lNk27WTB?zCVgcV=J#wy+qUFpn0lRRbL{A z_E2ux8s@@c&Emcdx!=}#Gx=LG4{p)!5P2eQjw>a1=^|hnP&EVt9&Nmi44pR>#)9|7 zlQji*yt)F^5DN%@%#iiUS`H^|20;h{i^<(WBD6-OM$>4@Cz#nl80EPNe*6n7PBM` zjgP8b7(14U0C`G4F;oBqLn44>{-QVDlYs;`J~fbd^&9WeKw`bRmUA{=?i`$OGz?4l zLd0}<$^vDH4MJ5F(R4}Nfn{8z=H_}3=3RmaVTT;T2IefWWFm!#kb=AEj$K84>NM&8 zpA$Ch^y?&_n)Ev)bs^!z-)F_3JOk)Jk|AGG1zbg6d-1e5@sRuO$wk7h{QB$smP$Oj zNUS&3a)9x@FYTVSk4-p5PF)qx!$p-D$0}i`Axt(S_OXgCv)jx6KvBe8n#3cllZAc( z1xtsJ+#E|gQz}9E2b5(TkkSQyiKYl4azdeY2;eY{A*Qalzsfjrqs z5H_iu#A{00xv=pEYUb@|Ubc=EDsd=X0A?o*6T{vU7IrcAL$-eD z!_0KCCJsw`KwYk~$_CjE;oy2|U=5RC6~w`!*KAT_sYl})vQT)>a!=7Q4Q2zrvd=2m z7`Ow>9~xTO%3(5{VBuHYY3}}DQmlrYFEEl=>6&FF4e6XpuM@p-m5vINr9y`UqjtLs z4~fLngigYora6-=(Jmq7kFZi=!(JEHu4tp>%b|e8vI-6bENBE9AuQG@|o^S+g=f?e+7&*o>Ej?RQ(3mA3hmxH%Om4VF z!CJ+W*#P0N3T6W?9Q>iOwe_4P$rSJ|!cXGa2N=L&w23cl2w|3!MQDqyi7sn7r{LVQ zKZ(lkqwwBM{1YbvtA-y{F)b3pHb=_Ga9*suLk*aqp&h?r+-8)Jlc*p6kADHP`5*tH zlLjn@HRl`F%=?`(9!{S?1xo{ zsuSVX$3U(r!s9)pfiXNKBBf|tc=8z_E>^*3z~vx+sBbM#wZta*C()PbEuFUu4tK~H zf6-B`wKV!DsUURJpIRHEB_FaEq-^2)>VEbbbE#g#L!`YrMXRZHC&4BjS#@= zCW`@G#L62vLB=lu1V77;;AO_YRgdd^d6mYj?Eqix$dSfry*tS>$M8w#z_z-u)Pr11 zW3YIqysNGM4!?5uoE`=p+TA9suWP1)<-LB)a`4CShWmyJ1;{iDI0|!*kB$(a(732rnd-|ifX6&_40v@^)}y6FeLaT`wsW>~R^I>lRN(A> zcn`YHrp^-m^f3r)D_vqAJ3;8hZJpNS0djG&Cde&cinVu0QJC^~fU_D%o0_6l5Dstn zMo&Iazd*F2k>bf>fJj*divcSpsvg1Dt(=Aijwy1yEMY-08FRX)BodtfO&rbQQTY?= zcvJ)Tad|w^h$mUM*+ro$XKjnrKa@ZPy@tkuUy}M#%pmg~kPm^DLotNrD?zOT{(p+- z|Kw;01q0?;b^|YTJop1V>svVo`FT&o9moTuxu9JPR4=M0uu(`{<;eNH@7Ly%-wm1jmMnTX3z+;~{HCB<0Yl_~q z=g9oO&`tA)H8v_aSYtX@HME&6k+JPR&P3CHNUa+dM)l2UCql_ALVT>BG}$vvM7XrDXmP(a^C)_QJc;RVuC0WazDvXYV8&dBpINoH3yegaEnHE} z)^bDdioc~<#UBhRG-lU;%RCsU$sIE7Mi)i^`T}nRMxdZ-rfv@ryZb!fP3PRaYeju5 zzECpmNT+KFtpkP#G4>H^2OZ={C8zt^=4Hi+y@>Ar@yz3e(t$rPG1q%zcGH#mUG7?ZHCdD!Cza zNeBQZet#$zd#_NG>o@fqvo zZLWF1OGzE?1TbA232LYg(VEJlLF7qepCe@zH1^$*;twWlQEdqvWRX+cIFtq_$Y)-Jn%tBiB*7bbX09H;pXJHf?vpBLX;a=6qNhI2Hh-6 z^~YrteipvKS)@Wzj$$fC&N(!$2%??>No7mHK=LKTzquMh{1!+a%^Dh8d0;|3ykp8G z+5{EJS-U;GW%LER*zFrUiWGHrB&Z|^QbUf@2oraFF3CPJ0>NC2S_D~RlFO2sR5FGD zcs>ZxQZFV<$F1|Ewr^Kn5g@td51w4l)ylSMCJu9A*E~U(5&Afa@1;{{{bC(yd0^@X232pT9UKiAh8G`cB9QN?;9UvneRst8Lm@Zw zNLW^K2zdt7n*6YgR5R}`TUo2X5G-lghALOWD@f!1EdM8vNyMf2Kn4_nM`3zpHs7Y^ z6+iEQx?`7$#U&LjwyJ4D-cYdQF}9X0JahU?>7t09Gn5dak+MxwtXFD0M}PmDdL^#B zDhH82AapB_fxg>n{s574#C+ltBTAc;ZBIL0&}m8fJDqG3OpZQ%GSBtMI6AFcDgChh z(!`+Q>B;6VQ@EWGTU07rRiFwz0pU=|3)gc@zhq~T&Lske=wZ~6u382xyzsKxrJ|AJ zN#viyC>8mC{lmx~8oRZXvj)Etp*=Xb(?FbibT*wJk4QK(kX}(WoJ}@buoPfUbF3;BQjH>IaRQ=O45gjpXGuTxGDDAN>Gi)1U{>>ol5XX7 zBCYDKZ79uJTQ|i_q)Ht++N3{$$ilco4>(6hjN`F{kXaZom@IqO4yRqt1QDKukCFDd zyHNU2(4X?!6uH}!D{~y)vRo$%m?CQ==-Yd$Ayz^;-_;QQ(9{Yg^wX%>#pOv28aJw> z{W8RT?)Z$}PsuPJSK6v{QenOeA8y5%N@9oSei#p86iIhS0H>C^mxkSfH-IrL?MvvQ z9VO?%yf>k`zCJ{ms=-)OWGs4`8jV<*{oVUGgd6NCU_AZrKGwmDEA5 zfE8k5<^moXirhfhJ?7R^7jzXMt_*lADycF;Ma6th;-r!tsJ5UcSu}Fi>%OwgDhTqs z$nl4&R&rGs7F4yjnTYpsGlp@E2lli*nNo$SLz!2iCuoNClP=hq%(cum?+SaUkk6ndMPn77dlnimy;f1^QpTzVztbK&a##NJw6X zWu!ehM+ABZ0PC|VZ6T6IM6-jawyP%9q%=E{NETUxpH0K&5$^FegxQbY^4S}mK8lCt z%m}#^eBs*xqFf&X{(PdNT^2zu8a$rV`#GVuHdmyr_2}+UsZ?{+UQ)+(*dCK|V7fm` z^^9ZgX2$=78Cutp_7XQs=?GAaZ%Z?O;OIT!K+1NoF)+YMJerv?k#NKpX*P zWr@~S2CkwJ^4p>e9CMnvjoE@>^ejdHmuXz;{fQeIkA!Ao=W?(&Q@i_hcLM5Kd|t{nB^|g^dzC| z9iY?qB-{zfet_EO0ji+@;!V)hUg>?%>qzzdfZ-hvI=O*e4gnFy+7mt{xLi(A)5%vM zqa(IA4r9i3MW5;VmZ!R6B?R$Z!1zNvE4fREJ2>wI*Rj@h7DEQ$>Hy;fW&K5YJ}slt z|B|_hpUhgjH9Kw|$A&z^*3|W-Ib_*zj5=KAylAoEL98TEV=^tA5RCL~=%xmUbs?C% z_W^t`SC*7^-e50P6g5G$-d8ZCdibwjr1}FvEA?s)kr-KG9Xi}pjHIG5-G^yHF z77I)%eTYYW*u@=b6SMfiE6vFyXk#4+2D&t=i@{Gb5%Y0sfQf`Z*)1ub;=%xTgwi1x z(Nd#Pyc_V>g0IH|+|>J>@dPEaLYZgFoASL~`Spv^`leUq`$|1`d^?RyZEki-&9#0E ziG5n#$C3=|15E1hpTh0MNku)16dGKVe1xr0yojv!Wg4<{sPjqRv24w$Wt(M4)1^Ie z=!6Q328<`gdjh7i0$q#+jK{LKo@YVkgy--R#Q8v;KW-SS+J+4_=5F^+es~s7ncTsd z)n$*w*oc(6I%kB6P{I%-yeyrtIhWau;NS$X&!(W!YHLyHN^YPhd3?L_isTZHRd0PW z$7+m3Qfl?lr{~e8ZSD8VP*l%ANA!ks6hm5^1BOkNxd_J@a`=P9jc!5Y@KM~v8uYu+ zz#!XDG$uP1NbTN21(O7lp(wOXFVM?!Up;Bu+m%-&Vmy|;xjPaL@tAEi5gN0+!Pk_y zSZee`COC`E$!@UY;Qsd!AO032gbc_#ZhkKRT@jdpF%Sqp1o z!hR^WO~d_sj=+6k)WfpB$I`M<%Nei(96c%HX7GpX{j{KFs*%>>a$Cj;&oo;^^4srf zot%9WR+3RUA<4U^n^%Njsdi8wi%z82frr+F@SO7+pP)uKC^(DbW<5t`{$*~KztMW7 zv7Q4~H@{HWGb#^nkAv$U7--NhmYb0L(U`WksQE>e;PKM=O2l(6$iSS55m%Uhhk?L$9@Wz$RuSJ*X#)2 z7p&>TVM)-4m|JWW4GK?k^#nyKSO1zzOqTsUj@Z|$ISB<~rF4PIl>P0SaP$KYa8m!D zyDl(ED1^;L*u@*osy)sCnC0%&#yM0SCI5rxj0m5GRNeN&@)GEvl%ycCzlek4iNnKo zl|<3d0vsu2c6<;~UCmp`Q3@y(fEHQT;G=2c(j_mosIH#VtSwwddka$Hm>}>y9_Uq* z;R%<~S+!rZIzZaqSGHOO(R>R`j}!d$TFzum6uwb}S1o~4-VQi#+F%l<<7e5)HXAqu zL-6vx{EcSHH@e+S#>zwkY(S8!2_xOd+jdwwx;Rpy_$$&ENnm`OOXAFkqPqd@h)T?& zSby+5szr+#(GZ8k>JDu*@=2IyrHSfHEa> z4qQS73d*Sk1bXPVTJq*g?IN6zwB*fZu|y>-RXARL7hunllqp3c=s5!SFN2a)Jc5wu z9OsbC)tI$(GuyH^UW=Sm@?vP8o?{IQrmFI%U?>U$hp0wsA>?nLgakIijtxrhCEdY<)?7j}ea$>{gyQtCb9{VsIVBf={o=!81ul0wdM_1U*?KW+X7%qhb>iaa5j# zZNTIdY~kkTSBQ0pcx&qs(^CZB++UO_7CmJaS^DDGsU;@cjXXQw9-`Bv1wi#U^zo~Q6A zX2->o?#Z9PuDmLRrbpCuYcq#T>ZqX{Caq{PbwioonRd2d3-{QPrVOPx)X{JiOvX2X z)&{0+sKsH_m+7QkBh@|C5g*$a0`#L&lFFpztks*v8;<6q3U~DkhZ#(idS_j2Zr^pkIop* zte8n5A905Zt8nTtlF=-o0(n@*F{7&`$@5zJ;<4@3N_wP`?tNp{-85kUm7!g1t~ z(rmz}Q2iVW4|#VVlSycMKfy;qac%8Wntm9>PI1r$*4}h@O*@PLAq9`r z9me-8Oih>zmsO$t6W~$!FwW4+L`4WR)>bkS!2=a6^y zRdnnv70^v*g|HNSnY0BwOOGW|8|XY?Vq7?kAVoVNhlyO`&=@{*fcQ*sjfLc}qe_bY zOrj#=ugYwO0rPnzHj9BY&$3AP21&B5HLmQAz}HVp{=i;BC#QNo4YkS(G)GI_rC*wj zd468fs!QJG1M+Awm?lFw$wZJgu?_6Mf^B34^_3KB5{fbsQsTb4@JN!DYVf3B;&Upn zwBubgZl2@P@NIEZU0;>b$sYh*t>u}GYS%w9Lt+WjU55`b4LBe$+U?6oH^;IF@R{g=k)mlAgEIH#oI-GZ0LPFqa9dGFHU2Gnl*v$LBnAnubaepFu z!MU`#Z*TDm{q0}5@z=lod#_*gmOsbC;oFkM`r4`_3x7y2W78ah$7eAOTh?7AiJg~VpLW)Rz75sEIk;r6tP`8u;1B-t)g`+n{&yD8333G4(0)1 zHKVT_9}J0Z=#2=ACWC@qqzR^qkygxfKvQRSJ?Q3n z!=MvFaXKXHcL3s8{~2foAQk?2)QzWou?Uz6>0@*!XxJ3%J$#upY^{b_gSD+5b@{9H zjU11$*-eJ7y6+t6zlOFSkvNn?JlautnR%X#j(YFf3-+u5QVyaamVk@`FDV(^@}pGg znn;a-ioihKLrIgUShNT&8Wo=08i~7$AXM5LT!-k)ae^nA z!Vfmc)Ih+jf>DD@BmM}h)kZA`{W`d4ZcDT<*>Mp%|Dnf8Ufz!7reM5BqG@ah zCvT>GM#7VXaf#d%)8U&>SEJd*=)h)+}iH47sl5?znJC}H`$YAgI#$=@X?w_d~`hz=elT~wwgQTeFR`|wL?Ky)akjw zu%j5zOLT|u19TUcsIR|72`U97Km6_ghKE3w8iA2LQJa9g!jGL2pVw401+$4K9|po< z6?_=nS>qA*T(9MTSG1e~`!4NtILE9G?QLsi#6U{9&@D~)lHCW^gwx#z!#yQ6!NjCb zx5vO`sQ7UX{$@5$9a5 zZFH3wzhMK z5)mV@1dt5`tyHlj;>-PEYi;KK@cJE&M>uqSJts1Flo{_IK3tEeIH9~wMls=%eeijbeAY$r}Yva_c(J#!T@ngGm6JZY4Ykh&@wEuK6gh?Z6Gh_Hgj zqws$g;TU-m{0dLbrABu#AnLDSf?YR?`(kgWm0N(5y25}WRBol zC+#SZM;iP)r?O1mXp$+!`v6VDlFraL1&8Lma$OeVNGuTRzvs*}FgeRLI;Cb5x80y^ z%m8BqTm*w)5OYCxL}Dp82VLA%i>4e;b`gYEY8Ub92b4#Hf$Bz{`PJ^Bysr$9SsL=0 z^Oc;wZzmFGR`R*973PGb?M6uJyYaq`Gx>j~>YAA_dn?sUX|XC^(wsAlLUx*8M8D)& z4iMkeGI8Y;UtZ{%QA3qCGvca52p`e1@j1}zV*zhbfwX0Q z$h~K!2X?G&Z@i6rkkNvZ07^H2oIYim14c{Jcb0IBkcQ^~8AM-he~!ZuIVq$t*41o= zP%x$V@}<~X4POd(VtE94H`a4bEGL0=K8kOHB@7<|>0^6h%?V@qy$*zJp zs;_`SuZts(Xz#{)4jZ7wvU6B`z-XfcUTcBp(Ar84@n`n%I8cvOziG^rx4csW!ycq1c zE5AM(%(~mgKYDDbHmf<1UvoT;DO`+cKc5yUv&{JPZj|)X&RQnYX|HT3c0cQwR4T{1-C4@ zX0p!i9k?KLsjt`55v7SNC@zA#=&opdJjZ&XFq_eClShgPm5%_8e7Ky2v0g4&4?wCkf0dbhi{aAmW^zugW zr%zgmS(G0a*Ns-CaLV`N0ij2`u>#B!iy4nS$yO!DHAG+;q(+0#Tu{t4Sw}cX&85Yk z+#EKqDf-vJE!q4yM9`u`H9+Xl4%4DgF$T!2vH3fieC1GJ)Z^qf<)Rq&`vO0%Ne}O( zPT$q%5!ch?MmjSu2O2R&=+mel zKPVqz2Rkz09_Wm|=L0@ToN>kC#gqF5DP|?yFDza>a#FT7b8s>>8%%cy4dsfmAz51f zZ4QC9jQ%Jy<|ih;S56O}mQ%e3+FN{e!Zr=u$9E8HN37vR;~oyb7tpnlV6$UJX{Bf& z_;I{YAf%2Lueo+*f#8vXvX!Utc@Fv2Hg&&pa2V`vsD&y@9WJDh5yB_QTT*2kVni^V z6WvbG>f}_q2q)6dvE8jqN&)`FAD#t{9wdMs?$O%5Aa*EWOYg%m80vJ0k{b#G9TcJI zRU)MtbREsL^(8{bk0*sfXB9jtEOb2bOSbad@b-oiaP|jHYC$u{ZVF6R7=-sRs(cC` z6JkqbpwX;Rq&QG)kA}&XwH{GQeiqXn6U3;C=x6D4It(dUujqb;;>?Bvz^y_Qd)+oa zz7y`sD}rHo^qr_~ZRL<+7xM!N@#CpMp09IROqO zYGTy!?o25}#(YcZAxY{JI_5Bv=l2l;)QCBT{ecJoBTh4>z;F}zp@JocA5#iN$3|sU zqQf8bSzF7K`8+?0e@5wO%<%0pW=(?rrP+q_7)+2j8lep>J?VrcyBntU2Q7z_6#jzX zJeXB%{Q5_57A=fTR2*0F9bhO2jw$xxb7YfT8e^OPwnF-7T;8m7p*$gqKuN_cd4ZM~ z1BpLYHhLNOtKtu)F+73l5<+^KC-E zBT0YJRiOge{b1TLjejK-V9>7DO>)Rst%rqZ%{-3JGLCqmd*u(StgUV3+$$s2?1y3J znng7jow7n_Y6N82M}3_YcU~Y@F6CwVeOVVyje@1$wYiWvSOApLO0omZo1!~*-!-Z{ zVUg)B7&C5{5#ms;Lov_yh{c9$y1|R!SDoWOk4Wk;yNade>T`@5z6m|x%Bzx^{Xvwq z${(uk?i(N}uORn)!@!x4uMJ|lg_P4Ny&J~&7>VH!Bu_FxMfd!x+B1IO#QBW<{`d;8 z8pb({8?hq5F|i!t0e1{^RK2^(#ysvt!{s@~5HIwA`@;!qG^kGxBMt#!rfq3Is$HYL zmRrY>6;n_aBzk-a1F~OgUSr^eHAD7c*8|Nj*^##+Tdvj-Xvx>;Pwu&2U|1HtA%0vK zl*=0{Ai`yxSj8XwMl&rvIP}W}>$|rjpXv+%GAftVt&z#6l)m^$|#7NL^D%iF`LOwyWh3M8i3kHc!%H7q*e!T|!s<_%;+ zPCy^Qhg5&2sqF%YQ^6UU>JUhj1BXrvmnh}6f;EI67Y9YpR-*0o>Jx~HKP0hM%Ns%L z93SoPYf(+5znQAU;bl?D%vvfwUDh4j*eqVMB4- z=b}KZ?W2oqrE`hoKTKQ}tLaXgL*OSweYjAMrk2@PX9n-)FMH&X(smfy#Cn&t)f>QLX@r^s6n*BUs)j$8&L)~q z9#X7CUE{-bIzRsDFZ?mP5r2D2#XK5+AGX_f|5ybe`rf@@zl7xL?<>`s(zhQg26yEZ zIT)%Q5ypBRYPADwc<3TwYI9e5vewKdc0&N{yU4Y3?S_ngk zXE=UhG3b5t|1 z`bdUkdUR*EAE669=vzPo-QS=Gjc(l09o?*Yk$RH;t$iX$Dx^rO6*`|{keQN5(Fva& z`|Q2;+G}02Q>KuGhYOr|9JS*{L@B0x(?SPOOL$=mvd(WJ&4C*MP&Xgs*|i4IV+#wYLSLfM?B5vm&MLij+-w^F3+- z$5EKckRwu%4Xbi-&?F4{)Hb)oz*)e~g^UEDvOr=HM@~CGXYn@rQ6XJKQ0D}k}9|uxG3=mpsp2p$2Fa#8@0y|cSA@XaxzZ-{XhNl7=ViFpkw@P z$$|n9Hxyo@2vi~AIM+_0soAxvF0m<#8G>1vy)kX}B3N`;I*%ok1hTGRe5 zamoKpzfY@t97cVVf0F(f%&lZF`SD0lFjc`L!3C2?)O77#K{4Y3X6w_@&e?k)F>MCx z(U*I{#AbWh=tet7Ap$*IW6Wv(U_cW{XM>M$SPJ7AHi`kkdhj`Jg9(ru!@D^CaS`^Y zlEKWnJp$`q!a5JH2qFx(OX=+BY@cF;@GMl7Pp$~N!=nR3<6Wbm|CwzTr-pB$X)=|w z9@%gdNZ^CSWuJO9TW)MhN%Ao`PL8IIv5_KjEC;1uNkrvJW0zkE6TL4rW1&noo1WS} z1kN|c58_8AiU9pHO2$7*rW{Y+41^`0T$SMQ$lBakD%cbuCCH4_VQ*P=Tj3yPp)4dMF{>TP8s=PCXTGl=-Fdy~n{`28=xFnqXvGrk$|_YscX zwBx}qG>JF-J^ixu_RZ4Ux9h)btiP)_>J0kotM69r&hTiX(AX%#>YL3?b->wV#R=db zSMj}OW;$8JBe3iN`Bqc@O$gj$fDi78i~2awPMKAk_xL05LB>5Sv8CLmbz~>icE+!1 zS(*3$#@U{Wq6_ubI7a5Tz;BQ{$`u=!9(d}DDq*MK0>>kNbfc))nAM?anVQuUl!szu zGprH#l`B$dkQu}G-Q%l=nWusx)Fw4B_u1$mLR4mSwKtN!AdB@b&w1@UE@iHvZ0*NB z!F_U7?g5V=>P8X9zBuKZ>4Q2P^|Sg=t#CYsgQ!;AW>KA4O5CH#q0^mC0~R>GFx$Nr zNsXNi4O)iiHjW60zO9HI{QlqAC!;xm<0$?ZI$X56xV2w~eTq{t`(H}?TxaXYHNkyy zRieft`ns`EK=b%OOpMqhbq?1xN=g_U4ns-PNj)$t?3teaJ@yB5BJGkrC>gRP(~p|e zkW#L%ErgtJ0*q;<}414T05y98zv1TpCV8KlM8Yu{{Qtr+2x1sE%~N5GMlj z#iaBs8HemAAK)9s74$i}QvIuqM25+?h$~nzgM81Fh=PTX=miXs`IRrxgno#1m|X-E zD3?kmBu_RB3L1J8xZ^jjj@DOf2nmm*nPVLjtDB&&I`2RJ6{UJ~E7lA`QEiA5yTG!;rl$Wsg#UgqKV=zG9H6;K)JLQx{(rUU&2SSfTbV4^qKG-pV9=B;o{&Ky@}eKQW=yRs_wYBR?#i(rCfz|WWq!vnn$kvf zDHE1%*C7=a z{KUNg!9L-vN745QbXUhpb9aWUaM9gyID0NifL6$oYBX`v7 z(A>~;_|;%+2SB)Xfow;JETB8f1*P(qHwDS0>>S3EB?56%2}=Z*UHqZ3#7%*Uq-J+) z%21BB*{ISTgP{|N<1p&>XhISe>?39?g7IZ+ijG%||J(K`y4AaKG1H1V9ndzeFO37Z zkIsXDAKY@21y>B_(99rGc0w)PhvP!@y+>f7t5^Q>w??S~Ip-SP6d|VzotHguG}~ z?${z{Gb-rUS->DK5H<$C=CcPkjw!GlCe*6xU2W5@Y#8`4C{P-xfUu#M zIMMjtunH4|ISuYoB0QcX_z>%<=Q26MFGu@Z;cX=Ch}?Z~p3H%V{UCDIIyT0|^38w_ zIPsJn*}``8C$0~JISzm+1Xz@c4#w%8>hy} zD6XgY#KL+EM~s}rAE|E{21*6#aH>Vl>lUVu*2s>>*2uO(G^?dW@`X4Xwrfpo{!6z6 zbRJM6w_x)n7>uKtr056u)Hl>5DU}FMFe)Bxaj6G?$z2?x!jsK`eR4(O&EIw(l;48x zM~3Mv^o4>HQ>13sjB3|Z$;8^`TT`<>L-`!2y^mO&mAxr^|7b6yCP=&9j{ka? z8BI#XT1%jR2Q3Y_IwPy0G(`A_T%>|_Pgx3bO4&^H6a$23uo#J)Mg=_KEK_;xf7XgJ z)s9*ZP6#(z^sOmRBxMFWYy?eJ-4MsR$Y)ALM_Yp!TkJJ)x_5E#gK>JV))Ku9Oe9zSoQ;$E0g92kTs{&ewB-#qhB?A- zy@yfv?i;#h{&d|(tu{>Q8hbzNg6K<% zJ(&a334;1a80QXOBuFq=v0p`DMg^0KBPI&CO6pfJSJyV*Y=-e*oH(5Jz+twmRFp&U zmt=C_1zWLCu1Fksq?x>1D^OlM*gLJCEJLe4U{?_6eiYpF7WQ%E9E!L}^gZm?CvpAL zCydz)(+W!U6J!>w!x)siyMg6!&{A^RpA=6 z-)>9ihLV>9TzHXsC&hS+D3eI7J1xe=aJd%{+El-Tax2(hWexl%CPdNo@BcgGTM8Bz zP9@tTEE9Z%O=iaMnsY@98;^XGckc=^(&=`6uX+4lE|R~@1+b8E#y>$AJ}r|i@#L;9 zRl#B}h{~kmhE^N7Ek&DBQsG?SJcwN)NtZbU0y9U7nFlFOf}!H*KvnKJ>4tIROJIOc zvTZ2MM6na}=$^3pwrF1ChBEb(7p)fN{$j8j?mdp;=dOhFq-O&6!Zm&~$ipFX9Esb= zdx3rfHv?Z!aus|MTpal0k+4=+z+v5)NfrUj@+{V82pEf-307QG9QTvGT@7v*V>`TH zf#UI59xA=u8sB=>b)1JHCkpQvI(dCt$f6JxdZw998U6_M;Zuwdo&|&P$*%Hzb&$iO z1H)=#qkydHD$ZFJ$mG6}XqSj?Dj`7rwwiS8h(<<4!dOM(ShXWaZQseq0xNoyskzjF z(aI&wDcwmQqtSEuw2;vyYm>ROKtq!iCFC{4;Zfd$L(dAY2#KP&Nf9MMmT#A8x7K_^ zZHt9e*oLD}E})*|*6^%=;*%@V$}IT%@O5| z)0XJ5LUBT|X=j#TJ&~$Zuxu7YUwL|cr zBOzGR6fLzHC1o_t_A+ugf_nk}EB8Q)2Uy@0w%na| zb2DQ2qP2X|52B*5XjPZoljDM*bj^Ck)ivOoLArFR-Zb!_)t zmCmGef`r3qn68?OBxz%U2l$L2f6!n)qNPQ1BScc%SlS^Jx>pfHM2!o6DW|k@WMBuufFX89{>1t|P@HhNSMsYX>LPIKG zaeugMt+BD}sS~b*eTAzv{GqOEOUnhF@I9{KHbMYA#ZAdM1~Xd7tpB+`1-h&qV}NC) z>ljeYNh&=AZ(9(m+4joaA?yv|bHJ=ILuZ?T;cXwWIzbL;N;9suXb?h(Yodz8zA=~^ z)D*K$X`Yk}CtrpbOLIev*S|ab0kmt&MIipk>8ITf$VY3eW6l*>yOP=*s8$^#^|?~N z0GfAuVO%uXI(jZjc9LOlD!tZJ;*rJ$PNN8pWzXM})dlBY1*;2}PW-{DYs*E86e${# z6$J~jzJ=8Ri!xgIlen`{3=ua{142dl8&xB+L;^^xdT?7d2+-79Dki8Z%fjovY%hp| zmDKI`_49x~fOV} z8L@Wt4V9~ii(#je$Vn+v*xL+#Ft4Ghl}7l~X=P;#5iCwX9uQqYiY5}1Z6(*TQn4Fm zOP*UQKwB*EAdoXgS-NbW;>-Mk7^{N$h08VmaM`s+5r{7?+bB+jm5elRQI%3rOiS(I z!k>fMvNDBIH+SW1Gs26xZ*Kd2(ovJKb6-nRh>0Cxg%mZ)ORJd=lTTtTKJvP+2 zkA&M#YHe$R9^@a$0+RAPeS>P6^b9re2?f8}Y?wxw4}g(>Gsb-2?`mJ-{xvz}gIiMm z(61eZ1KVd$!CLoW+i&#SWeXcmHWwtDD%f0Dvhiqhv9|KIfH&FtwA}=YH4B1I=(w^B zh;<*hM&>WQC39itn1Z@vKXLsNE%JcCzzr^!XIe4#e-dWfO|H=0BeRSV))^#N$c|J9 z?Gv1n;2O>w(g)oj(q`8YGH(}2?bCHKU|yu}$3bw7Zc2dFWZ6>4m&e6YH9Rg{gB%;_7ul&SBno&1eyzx18+~r}z{ni)SJFd~!wLR*x1M#79A9Y1R^^ zfl~pQ_~huCiiR2vx&(3?vHuHBH*m_-*^TzT)<#2mTqh5W;+tvhcKoJx1IOu(lC_3T zCxvT1;~W9xVO|}h!x2EfJ^204qu>8uys}S?i11-R-TOl+CFvlPaZS3hMP!)1@*r{G z9)(e8gr`0+XM&(?@qpqGgB|kG2XH?muaSv^J!RzI=Sz1HaGtC;?31e!aQ-;CYa6Qt zd#0Nq^n#A@;9AI}Fk(EV*daQIKu6|g?(EQEYM44CHuijEYx{8J-B9HRk#-OvcVK`% z$h@R4MD=#CKcF!8p1ODW@th-kV6&n50tt3s#@8)b2l;a9Ab8eN-M3e7>>K{@Vz@8` zf=_3h=)RZ?TrM!NgZ^YHv`>u|6>kK*Cy4Jps!UE&j_5kw?|fd$I`fl|?UImk`5^A; zv$O9KM5_72Y<$?nn3BGKhmxH)HZzgp7zOnHMd=F2lTQaRvtAX}Zo?laTrjY1oi))T zW!Y%N&1CEqa!C`Sb{qdc710hA+KfAzB^pbH0Lm9Yy@7rUQ|vdjPns@i+>GnOraCJT z?v7-pO-Ckkg+4AIWz&Zzdk#XO3iceX81{z`7umf@Lr2S?rbjG6Yg9_C%z2M#9|V`q zSffcf;nI2=nB%ND-1b>+yEGRfT29xJsJlwgL$I7-=rR6oWGyGDvd@np>N*V`yKJPS zPnpcMQ+hYVm-EI_HJmqG*!Y8!mkLtI0`UNdC3Og6d0hQAsU_@R*4wc>HHMZ@KBA-E z4?EZ#G*Q6r25BX&d&!YzAk-aa9F#4i{JcAmk!0)ppF368lIJQ8Ky! zm3vRwVTK&gUjz;=_)nS*nU%-p8kA9^=*&J*GsEztF=C2AS@fVBhV=4-l9vibz>cMw z@3j23Aj|MeKi72hRsWoZX2NxjL_jOs^up*b1&aHu> zk3uFn;A;-~h`2&|LlZT=BJcF33GR1hWqKF_N_G(X2Y^C4rffhw#jxX9DA3wkRXD#M zokAJ~du=CY%O_h4s{L#}rz5IR>Zm6s1hmx$yKz&~bd}9w=ZtZt7O24pxep&?pT!v2 z&#EAbC3WGNt`j*g8zG(?Ku8%?Z~(Ep%A*6wY5|J8&pH|TVO}0%L}+F516y62pu+?F zsMNyrL96&Y_4g>eMzE-}G3{<4>8R92zO>TCGDx?s^2MrYyO6kGq+u!02JK2mztCb6 zwR8sPP+6C}Z9Lh65IR+`1#zL{PZ)Qt0Pn-8sblgzO)82X*($+(`4=mKvTm%#yA8(C z&eFnqN{_{9Vj>qyQoYqAj4=5oH5k*m)ptVn#QjDdcdW%oz6$D8(sosqi`>w!^<1o4m$2UAqO?nt0li_$i4`@v>CB-uP!6SVdPivn>^&LE)(A^$PgHCiq zZ-oA}?Ua4$;rX=7;Iad-udbvg*$hKcJ}@)-$f8ky5O67ss3@@w!ql?#^X$ufW2qYE z8}51b2f`K&6dg-<`AgV_pYIsEGqci1(ANi?-rVym;KUkQC#-R=ykrfe6@$`zSNuV%Twss0`diUJD5{dK*ib?<_%GSP?E^9BX_LRVVF#cQnddv?ItYX^@MFB7mCcI!FL$czR%l!||3CB*I$i5b>?j06N z6*sEy_u{0ah0?M&cMSVvhXK4MOAR?VpY9k-Mf<}u3?MDw7bw%xdlmY9#>2_OEn_Yi zWo@^1o-(W##Kv=Bu^$?@Y@3)MwcDI^AvQ`qogA4b>k9kiiUfnd3Vo?aVqj4#-tg=N zi;7vG2*73+2gonMEPBSOPMs_MK&M0_5yKE$RN=oR6n^U8jw8~x%q!er#KnmFT8;z} z3&BDMYA|o@^%UlV5-KWZ~tp0a-0*8EvK{4ISmIIS`2A4p#y~j$UKNRcBX^) zmN7VIW+VIl=_ zlm#e5$u3D7l@hh8G0^1}F~G(Z`KR3*L|~OKzmdn8d2*MqPp(L2@yDiED%!e|GxW$Y z_;hXEzIKQh|=?(4J*tDIP89D?r?bbMv;yK9$7YEXtYld)J-MG zR_Pc*ipP+SB=_0kb1UIxLtu5~=OE)VXbgnFY6WNv76Kk^CYIIVupG3K?a>IuF};ZBVmn}~%(6Mi5p^l|MFIch@x6*wV-U zuR6qtsGTARyM%Ip<}Y5Hz3?``EQKh8GQ*y4$vexJN5%5nN_bSbJIf<*`>x>5n&GUA z;+oC~g`eDLw>kq|+^WivQxRBj+SdjwkXxJPOFXSEbZA8a} z=(6A7UP4el^NrRmW$qR1Cb+@td1UCltwU)hD~`J4amZzgYJ0P(P1{$;^y2|3tc%H7 zfNd=3FT9R&Gg<;>*^e5VeR5T9I*;J*y8{-Ep;x~FAN-WHragqrbddt10PTEJvC;FlH#-0yFqQ80?X}kT=4eZ=gyiGVh9rv2ZK7 zKqml`VG_d?NeaCRcD&gd2h8mY!9`Sp^~zS%>F~>lBPmZMogVV%LX11{;P?LvbnajO zSJU+?gYjZQ;i(U+guR6eCXXO*5ra@$y3XVCP(w&%4I3_^!l$5`U3s~SD_!8<>4!f! z=~l?TfYh*J06Jt~xni0u=rSMZpj&cAZ%37V3;S0!GL0UJN#M49Sft_(Nzz__P``c_ zMBjh^bJ?`x$qIwCvt9wEHWw-WAl&7*KdXLbJn|caMXa4l1z^*1M|Y9i!%W}#M_}Hj z`3Qpe?~NiBI*FWvfT~98xgfmba&$tD%#+oHeR5Um4}S>mauJ>I(@>aeGcmKj_ciye z(FYm%EjR!gVeODqknvCFE0?*9Lf_;JAmNe#%(zWl+kAUCReWsrcyFMnOKcF4!N!a; zC&BlU+e|zYZ~X|c*(X<}mhlG#FBdu4&+Zr5ZkZ&Fv0CaU{QyWmW4gI}2^;8kXz$Uh zOrmpwkhy?jv0G)3mQ@I+0MV=fk*wWhlw1Sol)Pl{o$yqS%p5IVo$Bb((PFt#u)M_S zkAix;&xi+VfaAH&oC$cS=RyEhsyr?@oeu;KoIr5>v}kP3Z{Q4scfKmi8*mUMQ0S;( zKE-8M{CTifo?uM;G9)&S{u0ZLf^uL+QL{rCw(*Auaac707lT1&kG5*cpHol3FM#F2 z<4OK<9?==tM*TS)0KRs|XY(G)MC}OiR`~&>NSavYU>a}+!e|POB?b6@ON@%19`qt) zQZ@suzWaFwc%kF<$1`1CDZoe<50eqC+4%m%c2&@L*Gw5qv1#5#+6$Vl*mW*uk%syX zikfYrmpzWBxH3FzPKVz^MJ{NMb`Z;J1whkw$m8pDG3rfo7J~8PTw~^|L=JtT+yck$ zbCCDOwnx!5dOHCcOdtn1yH`G}z|{I4^{BBs0x z?u7J#nnPHK+|)RJ_v0VJ9-8pg<&Gu-^Oxuf@ypGig9{BLR$fu<*OUbK9dHy#r=buf zQ;jc6h~>E@#Ov37{&3CZ^#Ve(Ej%HST-&np&uk$OBb{$QIb{{?s97awyJ;G@AZ&8^ zU^nd~j<1M}Pd|u@eMY?^8Xk9(spcPS#)#a0^alKN5-MYrfhI9>_<1GptgB~c5%K!* z;18NyE*jG!=EThXha0LPpR*kba5@TZYT^h{c53OVNGs6smz^8Lq3XqN2|I9+JPlPl za{V|x$@Bs`kLr#xV_!D1{KGM>=fTDeFjT1&&sv5${MB=ksAt%1~MpZ;>7>8v!fw(64 zda|qF$6!ykM?Z!}(WK(6x!YF@`UA`&Djt&m$WM6`uZ8`+q&AG?) z8=ffql2A$U;4av57EsK7yYc6a#dHxMMKhsZlC|h0HkSpFHOrojCvO6t$CVV9d`)?I zm)8ApF&oPTL9rj-@z7zF1o?!W)hUBLEdu;wZTFT(_3C}?GwUH;$WYYmbwm)vDfl9X z3_Y!L30CpkRM^G%6BpAo_mk+F$99v@I887&$k@GFE$E9IMd`-;+5#xmHzRUzw$^N- zYyho2fkF8~54S(uuu-(dJ_CJrzH>^G+NL}t#+-)@Z*^FURZ2~be;j9>AMX%uSP`Fo^6AZJUKpb(dWt{Vh8iDtdET-A-L|Icf?7dk_K zz+GddU`BAnk^9vx--3g2!+@-HfK)pkEG_!7?8GW>o?@!6RIHw{4~-?_D_1# z+Cex z^tIVnmo;D6v-BkRPeRWG|6jkr@`v^{ieyVOS^v+&=ueDqHB$PJxf?NnTrf=o3N-rj zgZAxPIL-9cH_0W$kF;ncJis?N8XYqe7)JvzoxekqfDzMaLMwh9-y&5|GAexO`9r{3 zs><4U!ym}kD3ZANDiC%XpLni$#QPzAo1T#}$%nzs2M>IRC;+%GmYJs=I*7bG0;4#GWE;3L)9~$0LzS4T4-c zF+o7^u6Dqm*9*9tG*du9RTlH)ls))l9rpfMC6%j8^PAf99+s;ao+POMLN$X& zq^?K;;KLB>Fd_AHZhC@SLR-YutY?q0O?iOlqmh7D#y#zk7=5UGKrWBE2J`2ibhdoc z#imc%4#wy(xJm{f@SO3Skly4aPMV+H@_A$8DG)iGD~Ir+3=CxUyvhF%%mvPS;&9ql z%rTd32hpzoc_sL)Sz#qq@hv&{RK;H_V54nDU8+zlUO;FS1D|i^1u3VUTN5|zo?j7f z0NLt#Yr+%jSS=3{1)YHw=L8-I9&TGt09qg$jv+k-EH&W=Ge>d2e}x-lRd9*Au=|h5 zpPKX<$pfN`_RLfD^vomL?Oq9+=px&f^UVW)9{W6s%%jm}*it@XTh@G5~4W^!LFfa50E7H@DP zCt0}d4AI#C&FM07EoIV-HnVn&XP^hfyFVrBuN00-j7jPA^%ULxm${Ty{E;5vfITjf zyw?XD>>Ig+IwP!i;2UQw6IGx$suWdw>J;y()})8|&-}}1=BiF4Sf$I ziKH!Hh>QD|ObtLskiUbJ;9zTe?A7GzlWGgk|9xMd)4fgC?B*+-y(t3P-7yvDQ7q`= z<22SV;V9vjD0@x#QsZB)h8n-SCOl%D8wHrP&U{(tow5wXUAc|>iBbw_XE9(3;zYxx z+YqcejgsYi;griY`w9@;owjz0;6^m{U&59luJ=A9%x={)5SpIQ~655JrTfXw~|^v9QH5q0vpt zi?a95UqqMR9g07oa%HIi?UsRYf4V5Gy)}yOL_|xRT#Plasqa=R1`j7M3t{32*^p=K z5IZ)(gI43*Rwf+zgb}2TA~dL!H}+SK>(BvMYI3bLNV$v00PYkg7v<7EoSL_PiG?Nn zl^Xx#uSQ>4F7P(GSFU9f~Dj5TLw7=JL~r;xMZ(Z{f_ zJ(S;DgruNHgeFC{sc6^rezw*~%0G`lBnMQ7x8x`A$k{#= z5(zKh)X5<|MW4TPQ*V#6zoL7kXgIBUTQ~4M{Tl-S(_W1SS6j!h4x)B_=X?4Uf9G5b zR%Hna$e)~wD=jfP#VvrlG2hL2A^YU^R=K?aj>VH`zI}2Sp7y+n6HBT?ed2&=|?fcexq^#Tcf|K9{>>}NE`ObreaT0_8j`V zwTc|N$C`b;V4E2t_)Uzko#sfl@HguqFLk-j zeRkcD-=h~TeX@J;2^?V=Sk`DAR>;8K%4Y+&(R-7)2QExSW7%eWB@LL zymsoh0s2^4q4!Mm3D4gqp z5b1Rk^d*YahT{4?9zq#tyY+c?mL3+aXFr@_g1|&L@Mtq0;r|wNNFOKMGw#ya>P0;y zV5w*M@o_TKbVRlh-OrK%<4O3QfB_FSN!ZsMNm>H&D15(Kw4HRCnyG+fiX97q-kR1; zAn6WQN$p~aLwFa<6(B2+>f>YM*Y0f{bDsC8l0cIkRG_WMpteoFk{%PquF3bDN8C|X z)Dj+|tgytCE==k>8z(qE%Q>N`H>xrz88)7T@CloD@2V0u{-Dm)}til1JxB zB)%c}`KmnL{-Dp*<$|{3R2t6VZHQaW`wZ5&=d+aKts#bEa$-vOV1hV~mI*tV&CMHfnC+ZIDl6a7YauK7t;fF@vPH_|R8>Gne8?jUF=t_*4 zz&y+F^K;O&MdD)F#NtUbo@iOE3Yxm)4^CZOD~Og|)ZP^=vrWto9nH~;WD@qM2N5|T z4qS)9ULdAQB!oR;z%ftkM&WBwAiF9b=|0MWS;)2j#VP`Sxa8`3LGku0e@&B29xG)7 z9h~g!$s-ja2|E)AsB3pZ1wih>JPN+oc6zwBvIvqdJ$9bRjj8~QOa2hX)uI7M!^WFuNAI!BwP*7v1pRXxHfjm? z!5!-}hj};g8jc4MPMSmq03nm=hY;g=B3iP8pO4Tgx~hFlkh~xsoKr}Xspo{zCw1us zKf04#CqIj?YhOZ9;NdY4GBg6s@tvy5H~;;=fL31f0yKtTDtX3|8U9JS>@QUH`$Gm- z*9&&uoVnCy#9kYJpM5u)6fz`wiOsAg(Le^OEXFg-gnuk#nT_WZ6lzVT^m9x%hhr8L z@>$?FCShvfDcRJcr#G>BF&V%hkCBm<%vV#8&@P!7+Mq4PeFRk;%N{WbbdOt0P_DT1 zC92+dTM1EhLx((ys;?FaDW*z!jqRhOnu#5uh7_aa+d*Sp9*JEQe02Dbj!kPtP$Ll1 zD0%emu-=_TYa^-<5GnYxJrnaaZ9Q9<5)Z;%`C^xfj0&yM1r?!eq$hs=+Z(Q)vU%Vs zV(ZVMFd^e!krvXU%KK`O$fAwJH7Os1G7EXIZCW#hCwTzi%=*b(h0z>wa~x@yMS3|? zM<_+kG#XgQ=iFPt%}BXNFVa>*r70T(PmyGQUJ$HS1twhf2or7;l>1Ibk2_^Dt<(RR zuujI34cY>oR1qNcp?wzQJ{WWw{H#lg_op+>j2%g$nE}*)@93JffK9HI)-8(GGIVmiDWyF{-D?Se{7(_B#qPvKYe8)Df6LZ@@ zbwxon(6IV2X`{YD;P9e0Wtl{M3x)~Cb=|TIDqjNc6cHO$0TY+~fr)EH(waN4lP9Sz zYI8gV$$_vEGeV}K7~x{e9r19N>A{GaT?GOIKPbGF2Ut=vj>bAUWKA}t$K*+FO34S} zYy`7mHlagwdCx%K;6W9{wV{QLj>zx-{AQC~P=XZ<2gXAji{iUr5bYmUds z;8Bggy(p&B9BNuY1W7hmPJ=OH&jcW2XJ{JCd-25u>aAt(2w#Hl1kiF-V8vyBu;SX= z0%|TRHi8l3p_6hD8!DcJFgMlu&R#@#x8LV7(uTeD!B&#aSPzdYwQ4j?Q=7R!(c21^ zP6@oAVUY;-U>ie$6$iSLM%^@`RG~U*RqIXdbQ+V9nnSL&)X{{nWBb;e=H4LnK&*17 zX>`tZR>@ifefOUOn9r(bXUg%fKI~Zbhn%hz(Q8h3jy@c3ONDi$Q&)6l6%b_$BM@D6 zD2`2l((v_0$Dy_hPez%82+B1&XCN$8O>bD>Fg@RoFzxujj{(d*5j!tf(FQ*a$L#(o zDG5bu5>F#FMq6BZ2o2q6H}K;Nz+o$%d5iFcHp$)<$ zUkNP2H6oCl%BQd9i9Q%)w(PQ~FL`+~LlxxZEi-tOmtQLqr#>0Q@apJiZbNw6=@_or z2d*D#J#=9`Kf7WSBtK#EzH(uO}sb#1voukI7zxX6~YIBAcJ zPFTmdeaa8_g%EQw;fh8vCR@lP)3BV-CQ(0vn34?QVBhK7ZCpVookH(@Wd%n;BXMORbkR7+}scvZoOI53lehmB} z-&K-F(COU;S2QI7bPu$V0qWGfp4*zctWN+(-~n9y{$EE}EIJ_f0ClSyNm3x5x)sCv zFB?>U5z&12z*m|OG`fQo;2_&AFH5&?_$Cu0NN;6HMU1MYby;s zzen=RJkQ;09SDT?un?F2l2*kaM;Dk#(1X8V<=L z2Z$OtA%g>^?jmrPjg-a{tSet;vfl6r5v~-7JI)$D-F?oqnnRlX^awTac=*2d6*Re5 zgE+_LB}3B@+jLS1)`2=|jdPwXI=u(p2I5vP93MCm`#W8dJ@@m=AK}Uq_P<}~2t9&( zMYQ(j=pthIVKl9OVsB=sRmle%4sOvZ8lmeXBJ3chc7ii;!3aB-o|!SF)HoCuwf7*{ zuiU-sOFf=syS7r1Zrr1K{8|xvdIsP+lEXG^cXFSbD{hAueF=H-P>J4zR8%;+F{25R zZ0^XV+%8TtKuyYvok>D*DKi%$EsN5`%`MujxU^c4`IA2Un~bToacAqb8EcXkTw*?m z26)GwmaeK1YWLK?S3&~bO1Bk%@{(&s=Jj7#w2)i5j?!@zHc|NQ@GVxHYEtLHIL{f)kx5E0x86d@0Sd-BAUld6!f? zf`Wy_+twD&#j_$YV`47yv1eiQgq9!|B-}uNe|7^YZ4d=+?R$@F$~HdE%{;=jk|NLu zZuqTi8{kV^o>*yA1uIDAE z3Rfi#^I_$kfXqpt$Z^n>CRd6sQ6attQ&g#J1+8otJw<;0Sq01bT2WoW zOPXTl*+!!b7!Ph!izi6e$%oD^C-@e@4{UC%=A~i&yx%N7eoHBC6#P zI|2?Wfdl>kOp)4dHqQEKO{+v=ow6*Poo4E`TIR?|4;CEfgo)`U>Xy>`B{$@P$=>Ah z)6QrF^u5h{8E_%8PE`YLUN(L_iSZLZOBHC7EPnj)B-V@QRcAr}nuDq%Pl5tfItC}5 z*ll&%n@)OCyxZ5S4q@dO?+Kb6qI#4zwl4u@&`g9#VPRz* zlX8s9HYE#R&e4;~zJ2oRgLf7o{(7YKrGlAR^YhuK#`b&IE->)W?+)*3Y$vfi!p17g zC-{J~QiecoR{l^5_1PUzd4tMoWB>~QyngW}W_QF;zOFNZBWI1m8|Zz2!EUOOi!S9D zizyqhohgin7$v1YLDSP1g#wzJ&5EQJHk7V@=B~6WvoW^J__$~coo_8L1v^sK;1h43O&Gtk%AMx=Af;hJn`E=XFV=!*Je+suOl z)(s@6;si9Ll*g|k=*-6K@Bc?k$Zn2)|1+$v4$LhGU9v*QB@|KdawZ={izhn)`{arg zI3Am;B3-@~bPD!?u!SAfoyEk1Dkm2Xx>DMr$7BKu^YPtmUhh=FOHNgZfllW7iyj_C`@>A$jbDnaa0ox$g@*yCMbvE(E(9m~*Ut zKVQKxJ#@;e-;BLsR|)O6h0t$If5p-oUQtL#9c?0&yHk952&~LK1YVQQVRwo@HplvE z0R&@NV&^*u%!H|qq*vxQj2r!{Nv<>WsbdI>02=trF>BB0^{RQ!WsrSqzUf}*mL=sZ zWep*@${JK%`fCkcN0VO?7R<6>(FwtA7%FWKeCCp^X7Ix_#XwZu=! zeC|o1pKw`C86&Tr&pj&iudfv(bSKO!Y{cFYEtPO)?t zLg79L#9J{a>;oOgX-G98XSo6YJ>~Z%@g)K)BX44H^c5b`jFhW0`YH5m^L3qp91%ZrxjOELk%VTx@6rKKO zp;mlyMG6>?1_A5q1^Xl|3JdYP!7>-5m<-KC)V~|nKO(P%=vL-m)c5ZeYNcbg|*5mwvIs3Ndgp`6Xk=X2qZO>EP>Bhk(&cEvUm z^Q^9a*;P}vl0~22*ZEgL`QM$tM*y=(9JXaTuSi7$Nl>A(?{~d$*qYFu#qgcu?zG+% z-%#giG2_^*vl2KL7LY)$gjZ=2doK9`@jSGh#uvu^P+b~dO7>GEB03;I3^9!;M_4$) z8^&lM>Wt3xf~pl#ig+hB*`?0mHc}U*&#t40@KH1pdM!4qHTf;qH=*zEE842y9N?nO zBU)OdC(A_5)wXlNP%C4s?FPZ(c$wGAbDfJS?kCnbxtnQDP+Fie6@{NeV-m|l#F>SU zo3bEt&UQ7+H^aDo9Cr~JB$R&tm+_DO@BdG&`KES2?ZZ2z-!Vk8ayzW`^$4J(UpAzk zVn^`2!e_lvmBPm(wz^T^gD?vLh}4?_^#^RVY>i$vNDjhrH7_q&spA}xKA}XY*&U0| zbV`2=;=s|L^lr&%2!lzIFuq|&dKyg0ZgCfjar>jMh@68&{abu_sSd7W654nI82h}4 zse)O8OF|xD;3De12#&NywsA3w9>vAy;Y|R*k}@Qva_v%&6-I6sgCo7iFF6asvF<0J zq-NuCbJK1kbS(J*t*}4%5SibS4q$^dZZh{_={a+N2UT36>T- z+9a$OYGa==3X(Q<)CE)8AG= zt0Z?`Ykp+?&Uagi94Ly$L(^Q6z+-?U9Hwm0JjGO@^v<_CJAY{EySD}22Alne^Vkij z40+yJ;dB#s^}W{Ry1lzGJNlA1M~ir<>tNd@d-yp<%ehw(h!ed{HzU6EJ-%D%~XlG6{?EN^^c zB_w;jKr1|nQg*$F%#oEcVXP$_ez=Iq4#G(<#*8OdT5jU3+`WopRHe~}-5gDcQK%jr zI7R(yHhsZD5L4%~RD1Gm5Z>v{M3I^j_q*f({ujVn_$4x@tgf?UF&_%Vl9@?y~Ue%!6fr zc}2k`}TG)rHBB`At5V98d~iz*jr@{m(E6L@q`YG91vd zUb4YOFd$t^P$am?9fZi>BVjXvZ@>mAF>yisH2c_e_YgYa$D2aO;IQrSr-$UG6Ud8L zEsRh>`TbwQA7vwF*%$U&33~;XUi@LN?}}7T$SEDGC>asX_^6~9gY;GK%qBh<{7HjA zi3cZyc6i6yC`xjD$mqLdhDWQ?USbUm%?j5I{}zHz*lf>4>U0r(=T9+`!e;y($PK1S z*;w&pkwBJNsS2p-(IVm9dV!F!!D2>9-ioPmz5vC`XVT`u;-G8=*d9oa=8v{xaO1Mg zDpkZ*u1G+@p&{F$t@4@7z^`oZL1a~{f`EO_~|`ky-`OL+j`T8wljbgc>WXP!$I`pfrO2gq-N7X z`H$Leo<@E)C6rkp1+8?G2oht}Iy#f>d-5XfQ`wO4vW>DrSN&41Hv%yu8tkqCR#*N=bA}Cwb@6Dm zitS%6Bjd)2AHSzG^erNYcy~!XLk@Ps7YthoM+FNRkB$oOia_|yQ53mTsAEPQn=IRWE;16|C?{k(@&GOxReDX4o~=U*qA;6w4piK=aZS zI4hQjC9}zs>=j-HGxO+j@NT1E<;iK_(I8lXkjhfyMvV-aPD`aE>lGJd}T@?-KlP|WJ;cheb zSxr|-*`xGifUv^$ZdLaB8y<1jB5A8dxd)-*v)<#L%cv_jzlOmzHwWG2gSNJnT*V0J zp*b!pV%?R9XJf6@qFtWQy*g+^ERQB4d$o^mfIX&|5yy9(UNT~Se-sP{X1pUNQD+of zmtLat<&i+lkd@hFzj5y$bb|UrhXikopl@g!3v?%m zs5hZ2b`oTp2!J0d165>M2?|lpCS@?vqOd$LcJaIv0P7JS(vN!x;r5j5HW8qKjb{|hi zcCsCcu^rx^^S!W(Wo%SOh^Bfu)&hUuGLTJe9`4r6?-?#3rn^L$-0bZRnj1Qu^+M|A zJ`g(ys&POJ7Ar?1^kvboGPh`Wb-1hHk9fMVTrjS4jn59!A0ZmgTrIYIGeg!j>&;|? z<{(%qNTlED$B%mzpC(q08d_oOcA6N6g`oh^VGvW_5BdSwWGfh1eH)F(C_gwCV)^;N z%(wl)l40`{zlLXBO>2#}Rk@n{p`;tj1p_?i8BlLx|Kq;Yg`*4VBp}6{#ZreDJD{wN z=RNpF7E)L{g}0=c92VB3YeDr2+~iKXoGksO9)Vw?_VpA(0LvSp1UhMf+*`j9F(DbN zY;(N!6hDaPg~&=(c&H73sOiQ^!Sv-gpe3%iCW)L|#3aK2Ea)muXp^IQ*5znrM^a{b z0nEyF$a#nM0NHXdIPCKXe_?uHl=FB+^rFXw%uG*_t~g?u+~u0n;HDSQKkq@7j;`pB zP;sTm*+d1}r6X=JjMML+64bV-f?~{ShrD42#Lj+@TqLkkqz?+R`wfq4VfmxUUDRc) zEXU)67)BfeA`rK}(7lrBYRwlMTM2Ir>$W_4V{EJxR6^;!hPs)ngp3e3h&$vVqs3D` zvuNz1?JWF0`SNQspP;ZL=!us^S+m_6CAS0xRT8uTb_z$zN&LQLCkki!EPe)Pb4OJ3 z{_r*uJR4n0-XIFNQf=#tNa8b_-BEC739-`=*3p@R8~@WyfZERZm$2Qw{ikKNj(>4K z_@|w%tuN`gKTs^E;_ zGLAoJcw?=gcmI6)Vdp)&IJRiP)9uQilSdf9 z){hRcS0PSN!B7pIZgJUEGuwrCg^dd{XY)2IxpNT&55c^3QrdrJ>m|yHN*yN^XPTdz z0R|hC<{#Ri(^NYRyY@Y$UnUE$pVw2j$0iL5 zNC&Ux^7@xxERP6=HWY?f>uLTI#v>fi$8@H*{oq=|A9GgUj9JVXsm+z4pACr4e(;gV zaH)=H8=%4p#%4cgm#Zu}Wlx3^_Q|i_wYvE6*nAXmShrcgxp#4-v*h~6Kn#czL(xPvvogV9+aC$dQ+J?fOGU@ zIANb$kqGhFdb}$D#5B4~#6M}zxvJWp=>W%4WlX@p=CoA}mrujX&>6_OPWF#*STWY* zPO~qMl2*!5VO^XhQvJ|sN!?2_(Ndv_^!HetdP#wPUm(H$FycLTR@oTxOyj&{MUduFVI?rSaGFTDOX}=L z;-uC9MgiRn2(yaiS9P#}Z@p|nR5lcsX0vAffo0=;3iVIYDKt=XpyI{P#l%B z30{@FzZW8keR4%E8-ESX5|{Pk^{mVx+ECgSY{(rcR)tF4f$NqX^5=iiNgzlvBe_Tr z<Gi?Is8!QglbQxL-XB@fgjKsRal@DVx_`)#E!SqrvQr5BIaQ^l#(NRiY-A| z%*kbEf0T_SBJc61@C;Tsf629-L2PCHS8VC9bq_J)c9u`+3!sIgc)+4j1Ov=Hr~9~L z2GZJv%2As)nJ00~yxg5!^YT&%B#r7(aAP2H?qU9na<7))Hl;il&_<4?j8EsZvD0|S zr_+RvqA_!eBXj}u?dc+|oeuOPpgCiYn{hC{j_6rGCT#yh7J>xd97fjw=iCo`Fi^J? z?SMmK{;aZT!jl65F_$?Iyn34OM-f>1&U}U{YK$T40ESzECo(_Us}PMTL^ur!+p z4C?}k@iJuTp&5#6#h2U?5=b!ZkKWW;!#IRhCJYAJ7{i%z5ScGCf>qfZRHSS1=uxrs zcBLSmEODUhoav3IKy}6y7E2lNXBKvblOai?=&H{gD(>klHVMM1OkH#NhiTn+g^@ro zc1GRM-T;=U9fR5qE})MYc?_l@aaS^8JlPNsF%_^ONHdT>RF#M+fK|~J8eY3;`*A!` zWkZ3T%Bu3Pe+x%(-6UvWz^hocP)Y4g4k?8uR4M#h5bSL#98S{u$=$UV^+_Y~0G&3K z%eogOeMGN}5TJ;u*Kwc9`0r|;N6dpm7Kq9jEON@8Yz*v^tCDg2A*xGni!2sR5eq;y z-3fFnfy@8}kaMs~Fdv5Pkz1GT&p|b3V*o@-i$$`{y~&)c#++?OpN-h!(L!-SgVMm*t1ygo$i=d-SuF2pghYGWTfnBd7S! zF;8n+^27xGQ9XfX4>8W#hcOEijQJyl%9t7g7bC)h=UH2z&0Bc^D1JT-JOB32;gygvO_&JW=t0wVu>Bp?8;CutZ9uu)Fas)dmNUbdyjx!GqW>5UuW1ep z#h?)EVru3Xzkpn(Dq{fHA?$LZ!nW*8U8?Nad6M#9eLIu#f5r9f^da|&juw$LjuI;@ z9B-Gz4D$==ZXP{0{})-{X{@A=eFy%yE!Qo&59I=7E^Vo1^Gw($ByZIprrMoYE?`q> z*fkF;Ydoh?y(|l)qp|O4!a7s$fAxyOqwT}e+afZ?Tu)>sLX@piE-Xx1HNnKEe#zzm zJGqF?;UPsh<>KQ!LgB&Xkwr{g2sttFCq;^o3hC6D9!=>5f1dTHNc2Bz63Zu71TbIr ziHdF%bSKs@LjUA;(PGhpx-!?#ELiUGxlFJyg(Xe`C$3w+nAVSz0b<(^z$-VyV>y<& z9L1@?PBt60#D5A1QW|aWCH_y$RY3gT-7m}jSfWb+=8sR`4Bu?sL?%yoA7pxuDmOHn zDIQQxr-+S%vsCs0mT*|RV2f4LIUKwn;0>Gl8N@9)C1SlBT@wf8bf`G0r^xlcOo8ap zh+zqe!Q*G4!NjgeGVfLG-P;z;ke_Bk^&Jn8j;x99t-oZ;=iAdEbg1^d*+!w{DCMCm z{!oaVLe11ENtjsFOcU!s+*E2mZxFq{HJtkSTgj_Zx%YJ4^T`zf3YR?sg-eYB{f*Bk z$3;e>L2#qptM^e?=#zFQPONDTE=!ODa0z7sH4-xxo&i)az42N=!#ijU+ED~AlAV(%+5mQxDm>lcC^!N{eRf~20w$s#6bf*BGqedzDw3Otty z5Xx7vhx|n{>a%PjppkH50M4iq&6RRi2qXbd^7i)0RjKSeqKHc?1$Rb9w4#LzYFkqf zLI^bIWA@?^rB~QRt#*V^d51N0R5s<=qdWWFxN^}04h#Zylzc0ooes?aOw9{+zJUuF zp0Fi`|1l_0pL@m47hX*Bi{teqac`eomD|H3dbm_1YH=voxSo}Ej3qL(8FX(UBF43J z1@-1;1jCZqgY3hGx8ZkgjJABl+-Uqk%rJ=Rg_J3K0cd28yQ4WuhLNg3w$C!j+CAK7 zHMk7mW1zS*^ZRj9Pe|AboHjBSa5S}!F|>9NQE~sfVTWCTzDb5jB|R-_l8YWQ&oQe< z!Bts0D5CeiuBZyg>8E{%M=*40wSe`9!p;%gOxVFhku0JTx4KRlBBE7?9UA=^Z##}~K3$0-+%^cC93{L0HYq;GJu2Pt*rwuVG2;{2vCrGo z)&$`vu00@FX7Xc1_TNA#LnSjBfOlAd+`~SkY{}zEYoBzsu~LpDwd8n7Nq^K{$tspho?N@Ix zK5ddl;#3kW-;N`bOWHC8-g~&22x31O(d=@S$+>~KkJIfIoeNSj)0(MnIqgH zzagXXQd!;?YW+XI(Ow2b@o4G5=@le`8R`Lql2)@%rj4t5)n7k{aYhzEPMvQ#t@|pr z+y(U0Jtpb3<0$<7f8$0ec`NwR*y zo(@b4+$CZ1IjPY4Od&8|YaB)$-p&L2aRF6Jn564xqW@x#ofv7H+d%6F8lIF57f*Wp zgi8hV_+3%j@JQWUS}P#?X0<`>EBi@2<=3>rT32SR$&JXY^0al7aC8~8C)#-4`Hl}M zn2D=VM?uFF{WED(0k^q2VjHBERDK5SoPC0w9xgyEH&V0=!_HOE(GJr?ibdIw@g&br z$W%a{KMfh5AZ(!*C9FRUM2DcEPCPyO;Z_TF!CDs@X=`)}oY23895_-Mqy{5r5fD}p zv`8K4ItspW;)aW#!vy|C=I~FF+<&1;z$Xg&uAq)P#f=&yOy#_Wp4kHOWf<6DCJe25^pC?0MVS$bf4rMz2|B_L4thUlc%Fz0)jJ=aQ=kE9`no z0S$7(0(8Wzi$`Zhc23J}JyRTG$?S>_{9ml5_K6i1k&xUJL0C+xVd}U=aFNml1*vxH z8M}K#u@o@)igR?}+>TTye~Tb8|CWU2tjj6!4wI>~^e3{KJ2xRW$xupyGCoYJDed8v zJuOdS@2hW-k*`YQ>=DyjUM|qPr7*hb>-gAK=;LfOv84Ke2iv}BC90WVpWO4>bwkNa zX^MyZ56m(@iH&@MP0~0BLkcD9!{9~lGM3b`%W)iF1RizG5`bxCpJHqfjRfUM7`J40 z_N2>B*sP_{-K$d^J$eKzEf?93%u+crFeG=ffbidZlo-w7knscJ+OuuWY_ZB58X02e zE(3^-&|mE)Fw&Z>YHkM(8;0>%HX=OftrHRNQUvbR+jfsi{fO2VRBkCLiDcnwF~?&3 zjNXHC2?=IaI^+)g0Cnt;qE3?7RMO}MI3?@@$gQnV-=bC9t$jj+-)Z)L39dR!o_Gy4 z?KpmkG_mmneC=7V>`YAk^@GAAhjF>cO9U*vixYL& zBI9z=JbO17b1@)Zh&%hm70n$F^#beJO|ey@?)BwNCKxbNI82DBTx^&!u7p97bbnR_ zs22Z)DVK%d(JTLXyYQ?OF;hZ+^-5yJBYwABg#CYp#y(?p1b|bePxD+TO24jYX*& zXsjD3CwQ65BgYV{rFeM@9ZT!?xd0I&(H9te6hv|nFV~?Uq3$cMR6v~FfxI4t`j?j1 z3nU$2s*I;+ng_j~Pko!?aomm^P(_T)X|-(CAW!BrE_304?E28G-nV4zP1PY|_!LIm z4ov;y)P!P^;gyvnCRK%sGVf6d3m5_uPl-l=$w&V5+IeWARm70BIOYAG*SsmZ+b6$1 zK(f44f9FxroiHj$J56~3T}E6oyIQ@{rnlu1t7OlnH!$n^_tbMY1CcdpO|X8pl18Re zBWV7I&Sow^PV1M>evq2}LX@n*-b!W}B_vZxWSBLam0X#-%%wO~;>5 z|4+r3k$kuDq~y`I1EFsGYN>?8aeI<>w@pD@m9_OVv`(%BcaDI;VW&Bn5q7Sr|0 zB9W7WHy)JvArhnMNHljFiZo6dQhQBKB73=%MRFYeh{o3{!a~RH-R3sGt^@L?h z3AVjZTP&-zt~5i3krb!9H%W10=4^-K_9#MI66R;VeYZNR1HkJdN@O}zwX#afrhX1oH2BY z7CtH-m`!ofoyknngkhikW z?R0R<;1z3<$NxEA@iIr{a~!@=z$U!6ac3i~BUebxt~5)7GioVEieRyBIMvPbeFi&d z!PQjOMuXYvC)yuS<^QMyAg7pKU2^JTsHG&xDvU6LeL>yNU+3n~o+6O{3{cWOxgtHC z#|4(3N*4fU0r31u5&8R%CPePXK1PbmTxY3cuf}=`2zIH%?Qh;1j$22Ew*ip&^ zQji_8m&EhP_s$7t^PaHht_n+AnXwZ;;c>ML!eQ`Dlfw}>Z%&n(vs_~?sPqeI8V8h8pi^oxVqe%4SE5_i5 z;pN2fVq+4^r)2VvVqP^_g(KzUdU?C2AayZSb=AXBz$$Ro)suQOS3QKNneW3k;r_vP zr*V(EUFl5r<%N#w`gEvD0%SpElz}7%X2{SkY6IQq;=(BzqctQg)#!%p%yKkMPTiB} zzI}25U?%pIJepqmo%i#s}Rp{|=k59uhD@CT+!%|32pRBzcj$z%Hof8_5`S zf-WiNG`I@GTj=p)A_caNYAFoufhoD1E>xPnseM8A5abiL7x$x=R4Q4M2>#D2)n^qw zm5|qWnaCr|*jOr9&}yB0I+Npz*?Buf}Z#MyE0BdTSm|@7a=^(tk9EH=` z=5)kxq2yn`CuMrlK%)YJQ|@8)i0?I)3-G;NS%SfT>cn^V$yKEN^v9Wpa4rfcPuXPR z%1`~I5he--OcqK%X{&x14;=Ms1r(&*QF}!G8bzjD+p5#XPIsf(8AN!x!hR5*syTEF zTANu~@DOJ%pmk@iwqswPA@6Y%--9f@XlT@yT1=A?g5Ns`I=$c)ns1QwYbRQxGi>WH#__m* z(qoEw!@M1L45UxX=@Jqvo+R?^lPdy=EO`Vw8$~?utxw0hANE+V6!@hEKrxtZY}>`- zMhn^U2%(sZFnke0LM6Tn`ZxuGfK0+0T}D{f?*$`zT1ncCx(I*-wH+`uO&f`!v!Fw2 z0%%2+=HLDW4@Bp&;LL{);@)tq=7jbo3Q2reNZFrU2314K{^upb^9syNB>&aN*h?Nk z*hZ12CQStEB*w<734V?HF^Z-n2(_#fByu;nN28IDlx>W#Jkn(*JzL5hMwo={$Nf$n zAtI`k*v#NaZ*pfsPdWJUg9w`?q#Z4@I3--vtQb*?FnS%_2Gl^7(FTjAme(aUB-Y+O zNj?8r5yU4~B?);1W*aL7(`uRyrY#boI$`i)WGcdEVOU8&@;LOm)q~nuKaU(UPkn2> z$qa=?glXk=fbcdRJhUIF1b{p@)SfU}iR%1AFsWxjM_DZiQX^!)vIP@r`kwMkC3O0& zVDhQG-YD8m*-cCWU_o?MeJd!0Bq9W-W&EpLy3r!J#gv>TO)tc;JryE0lp|}D?1!GD zxBoKjs84b8#%ckEu=nBD=6mdr(BU!en=1;K|Mw$S0^yUu-i53-$!B`lw4~)O;$7~q z+HgwoC%09WisO&lr{2JRa@iwzim?3)9f3zeVq>*n)Y$h6?)`ESQV= zR)AV{ZRcVX$dZJ=WNP=6@D*Y*4W5Twt--z;b$8_Ce1EQ>`*S2Sd5ob)7aDe#LkxVW z%X2GNL0#TuMW68AYJmlOvlBqj;t{%s&`DCwMhb)qY$3+o1WQU$xUZp-3!D>i5n4R1 zbFn3>(%{D%mkW&tr`dIuUa>}Aa8L5K%#9Mv!!^K~v4P44&QlcWpI3BkELUWr;8WYY zQDly`-#p!R%3b^LJ#ld#YoY5D6_-e?KT@?DM5W`+jEK5JtwhR~t)_DS0!ERGBrSDb z5eSr9DzgL!6IKZFi9M#f7nB%{J^*7XuyV=eN&5FMl>2?6f<;SrC*v^b#5JeKs-g!p z|FVwi!zea?f+GwIL`&y|!{(8ZS+>I)I0(h3a~NOC?Qm2(i{LbILRgVe4Y9M3#O?Zi zf_5uc#=>BlhRsf`g=Cjx8%2%s(Pa5jtY3S}Cs*Zp_lY;I6^s;|?#~GEUqL)NK3C6) zIeA|$t z00}Ao^ymxEE+-Gck$Td+w@V5mE81c)AI18~)-fu5XLn4gCP6pk%p@(I7wNUwaK?-F(rI`Z4b5da z9}U)*JHC^3>`-4+GRQ;I`RDA6Mudfwth7IXoVh%=OV<%kk)nU0I>IN;_O4*MnAhu` zJ;!!10h>oOdf3*DvCULBzgYUsE(^z@9Yq(oDgQx%!4L^q*1is>^s#tSbm!U46y0A_ z%)xp`pEBHyqHU7<^l7({cZ!wqePmlb;?6#I^=QP0rT1++aF46 zhg_03tS4Kvrk|Ww1jxHX%^lJ@I{v$922v?}4B%8tYH-#vyhPpbBw_cLdC+|-bT^84 z@!KafaqI}m4Oe_*oQ7j0J6L{>lg_yI*)Cwa$`^jF#?Ps(=T41Xa2T|wb#Q{Qc0~C*-02_zW_f}ETqeq$J3ba&Kz33@$5OL$v^ylB z1UeEmd0icfO?uf=H?PXFN90MCowSh2vcLKazu}Q@xUy6*T}T-pH2;XBXyDFq;ah8E z{7kSX9rU|~FBcLY*%OxgZ3jK=7DOX0!))V9%J>JF_pmo%!T|PXXt4&eCLHQ}yBy(k zPovnGB>U@!i$_3mrAUVM6yw5?6%pciw4Ep3SfJzrX3&8{UegvGdT|+lO)|u`SviqX^h9un6DY*r)8l;YWYJ>;;y*tFpw@`voXMe zHPl8wsBF^N@Wp^c9=-)IAHYXX|V zPuP#S0_7o;z!(D5fTXh@#EdnlIwTFnH%uSg=5;zzy4q}9U2aD4 zmrkZPU+;acXM5@y+QF&rBgfAM=E^@QDV{0*HVr-$( z)6`BLs7!2*MC)Qx<)ojW9mu7^SUSR+;RE|9+&7Yz6Txs2u!D)->^Mr`xPDYg+6vfL z5!HY{6FDd*qEjX!B@gi_D)i49Mxt+Dl~mvnMPvl?xUK~j1ba;o#3r@E0PgdGYvz3+ zul=aoqehvj;=@rDBZ95&6p@lRhMwfyP4XLcsg&W;2mao^<#7;A1uQ!43$d8pm*?C* ziBW_!t&H1@#1excQ_0%0!Q@FKf2|rK`7RZCggI9l#dN}MHf_)pbIg@N?r^)^JG8Hp zxXyprenY|QkmF7SS<~}&lygr>#m%#{_MOpE&&XH%h!C1hs+(?u`|%B=DwbNH3V)TrB-JH(K#HU@|6BpDNxI3M zyNO^WM?j?*?Xb0%LZ|G{#0*Snk>mcQF{gHa|F10dq)xhgWhrPfF(c;G;uKwdt?GVD zez4CAU+bE^%G7-f#xzFkuWhd)@_NwkCQT;oGzoKeiTnuEfAUHRrssEtstaggfrR>w^mjp+Z0TK$}(i zyge!GbKaT4{_9^TJ~7zUf@1bG>^fTaemsbow1l82bI0`=wZk<)6er^NA&zZZ1T@SG z>jX^iCkew{^*#s;cY$mucQJE>sr690SQZ^p`V_6##yS+QOy|@Nb=~-0vY$X}t_T*i zrjl3;K#~dD2C#Stud)&ZzLfZht12k*Tlc);6O}FEiKcgosCtL-UHu}O+(jLbGgFh< z9WkX9EpL)Z6XmK~c6F0$c%G(j#gKO<1EOv@;}BanWN6Dl6hbB-aJEDXMai14NA^2> zq2hShy(V~AHYin_ta}sOw@-fk{%GlKec7kr{>pkmX#y-pJ8NxGI>Lb(EUu6({R^wV zXA@O8=~0@vV%OW21t?ja;F&Oct|^j7j;Y=n4H(kDWe-o1YH zY7yeI@mVj}9prXthb)WmkW0s0#L#+XA9cf>BlMQaa@aK@=;Om?I_#ol@`FIE_47B` zEKrhFCm|pXlbrovcrk^8h*Umv07eDqT=B=N#5q7IOM7q36Z%p%Z=yggb2?uA5!{t4 zdN^@UlKJ+@70EU}8=-fN$HxTFkm|IpOqAcJi@)q5?In;7L38crH-P-OcQlg@ElYNl z5g;_wGud+@v4x2OKB;qB6UpPAMggHC>WZv7eq;BBZ*-&0Tsv|*ZlExctBK0>(44O) zL45nqz8@U&ir?OGA(V_+IlOBe#gYzDM zFXE)T;>NBh6~DTiD3e8%w9wI|1|Va8L;5k>aJuWxT_RFAGGuRO7sN;2ra2Jdvl|GMVc=8He&!;WT@1otp)b3&f=# zX{+3(ydyX+2FoFYDq@Q2=bQ6UNU<7ya_JQ+j`T00uWx7OvzvKWu=VEL3F`~)gk*xg zmK&X}zD!Lw`I>|27`2sy)YtbYu(`Y8WAKB+LM{Z6_5otf_CWK4{a_mKBe~t`j7<5! zu~@h1h?Ep|`iu!`A8i4D3f~^lR@f3{I0@@-pQOhBtf!q%uE>qylh9JM9MKB@wBJ15 zes62He50p5tUJFj6|&l(v@Irm!Gnm_8TwYyz+jYoIUz->ROe`aDvxAZV1mSMj8@1G zLQM^&$#!hl1dQnGvT^Yg>Hp`&#d6A3c}+ClGKx>C%0@vAdK${s%Z7sLNwhR!)Dsa0 z_JSLcn+n%qte|kS!c;tbQ%orieS{`-0i^N`vb8J>g^h@Z;zrDnX+?n?cE1C*XlX`J zE{$?ZiQMPL(5>OVXHy%xo5aRN(%ui);9%O*zIZNcr; zJUSF9YL3;WQR8;7=ISqtY=uw5z9!XAhh0o}J`EeT^&GR`g^7)#-Lw$Szmc~OHCPlb zNAQ(xc5oE<`LMb4#)V~yC6AT?t3}ew=}|aGAE(-xriNCCx=v5bfyc*nn&<+1VGa;c zP0W2KbB>Zc|5jnDH??1xXb(d``>azfdy729AmD|L$RoXDwTQ`nZ>Pbk#6@dD)q|R% z+}GW(IQ!lL&B z2(x+7N!Y;Zajy(Iq!AYQ7?s3VvKel&SaHUhcgUJ|>J3}lqj3$(t$6dqd-4#lPj=Dq z>iy}GN0@STrJzBz4PJ*dI<1w{P~tmWI}VzdQ9}e=a62(SQw^dlV^}yk@wVERODB8W zPw>J7YI8-2-a*kx7fL>{f`qPvTZia5qXs?V=OvTJQ+x+rsPOa%Os=jJIMr`%ww(H2 zx;7v5Q8J#~S8>$spGvcAMwd*i*lS|P0JswCCZ<>Ly(S@A+XdCAcz*R0oh{RIcWw?z z5Mf8p^=3S};>&}?ib+?53Q|z@P)?eQhtQK^zkPC5lD|h>a&@&JaQ5nln@8`n(Pqlh z=|sryj)@maLB&URghv!*XN=QPy%Y`~3fyd!OYtvL?%OUrzz6MfZ&L2c-!9 zm(^4o_@5#vF$t!Wl-XEW5C{Sx0thC6AO)@Z9cDE)i^*EnqBp(hRj=}!evy8X`5nIz zAQ)hzSPb4|EJ`9s5c(>whhO*O$Io$)=E7~jnjS-NGQ;;k7?bMF5nmHTqalQ4IP1C} zde|9Gf=+OGS<+&J1NEfLZ=YP1o4_Mfxw=+RCmbyAkQ(0U`hTDftGwbgU=F62Beo;c zy#-oa6d^DXgZ#A+7GDAzai>5LzD{Q1O(!bM!R3yh(T2?m%#!Kfm)8EeRQ#)wK72xo zYX#{8XQ|n495ZoO89;>))Iqv16;VjTA!x@f8`o^j<>~+iAKIei(g&~T>aM3{*b7I{ zYS{nX32RV>&$T9_Jb^c1Z7=HB+6v}!48MRZqU&;dxSXLUjeYy%iewC*8vCp31sP)- zhqeHK$`y{ODo9#hxc$i=lqisf!9Kt@pm;=*C6oYB|Iy}}*R;(F!A3^0C)tc9VD+wlO zz0^Lufnp~l>Qlt`IZ38aWl!-b!v4=f5tdg|Ugm2q5V-L1NX%RDad&EF+*9(9kg)K?Zm5M=SBy3B1 zO~9odv(6#?$T8kz)E;bM>)rdKuuEu;%Jz-Ej0U(GGNXZ4hZPz=An;hYOD)Q12_%7E&0-#&L!z?-W%vJ5wa$uT&}x0K`E z_b*r#08dw`SvIA3@)RIOGEafmPbnUO$+foy+W-u?SV6Xx)(k~Dt0N6;w@|a)CIV1q zv3ckX*G{Su{5akO?=9;pIB;zm!GP;U+9h)Fp!^~{L--T*XsX}wE8M!CC=p~BCT~O) z1qL+*zt8$DEZ&l_^VfI{xLd&^dbqY!p#1HI)#p-5*iJ?`uO%q&$UkNf-f7ugFyTj- z0I0Phi4YE;%_cX}4jileV61ImV(#JA*qY@HjisaOsCIZA$dqP3t{sjxV-GOZcz8&5dpFAvT-${L|bKa0}05ba^g1 zfhcIkTr_687#($^byyKz_95}4%fBwg|B7_)K6UxmmI{_WKOldE5Z(aZS$HM6JD#f) z$&oM;3w!r$W8Tufui>2I!zh@VqCkBcm&Q>B94{1d3J`l_g&QM(!u@^DQn+3TR0ieI zlUz1Po+8QrLeHs3^lfdqU=D+@%wF{Eo$I%d%pFSXp$+H`G8;e@pHshAQhlI9lY4lb z9fCFuG{~i_PC0+SvbZfCskkNjzHJ-mb}V}|PZ7d@R?oGvUKRAM;SqgXD^eh8Y2z)X zdZxEhHSBN}H^?=Ss{2P$F>xApLeWcU=XuW~zRR9LR!%{*O6)&kL0*i1S+^=2M)1t;kk)FJK8Dx$ZvJE6UW?J+-Do zNXP8gsw4k63|*_v5I1G09f)jH0Nn0&iKs6DC=Rw!H01s8dQ4y(-tsNfKBSdM3m06E z>F}BVw|Y}JAyaDSPm%n{DeeY0(PVsG&dYoK?UTg!pA|8Daz%;*pQ`_BD+QajX^hMb zbLgyy9lKq)9q726>}IM~wN2FE5c!zo9hb=+>#(+wKvmeACfAzx(1K}wflXPuEpw1h zP*tRUFG~jTrT0&?HL3#g`PBPgE20_9IHjh>z{6oX+wQ0NOZz@e6La^_Me@z4X11hYU4^kl*g;Ue;@Wpyz>Rib z`tU5gWdtu$({89L^8)ff>>&|2X{#QBB`B2N7H4ZpNmhH-X>8{Cwd%1wPaoD zOZ%T2WxXm((91pn!quXkB=op>YNu%t+QJT$L2jpZcy`>%a9)T7LV6kQU>lR#M|CtV zut6g}fQwfKcmj!}c6diFNL4YzK3755?jh`8SZ*CWxCTEBDqAp4C`rU{pNw&tfP+#t zT%IEG|1x02vQN-(y@2*{7D}E6wwSG~8V&58QqDm+zgW%osQg{6_^9M*{-&pU7}j<< zI>{eH5q_X&DBrmQZ5 z8gl3=aS);*C4RaqEHalIfZ7ie*M!@V|jI5+Vnp@r{_J?D^-COmpwv@>qUb*@Vu)lT72bj zTU6|5Sl_`HXJ^6LF;}Lg-M$KLrg8yD(=G5%08fL*I2a^c#4VN&`*sl%EiYE2p-1m4 z7&4{%S0BXtRJva;+C1Jl+kc1umQ`G-E}w&%)$2ddU^83r!1*2&4i7b9ATU4(Iei)L zYT@`I0WDP4kVw{xMujIWdy>ya3J|_Jg~z9sJ!;vH-ywzpg&6#0nrSXSi0Rc2AAHd%6o2#*wexWF9YcO2ct2jW37;uIfm!=G|#1t7s>-m5YQH1W=|7*Lb9ikl@@M zfJrwseo_+w&^#+hKeD>cwG|;s-nGS)G6kqF0eT`K6QFrSM z9ky;t4O%P_6D8q}0T7x@sZn>raT~)>I++FPNfSV4h~W;G31-yPqt2!Rp5J+hYWO#< z8>nbwR!jyrh%lABZqShTRDf1Oo!$WA6^|(4`bt4C9PjQTop#LLc)xC<;o061$X{mB zW|0AJkeQQ)lSUoRrHC`KfU7bkTtz4Qm6DT{lCsCX=Eu^L98d-Ecz5g`VaD~f0?rep zg`U9OZD0!i=J~(3B$&$&Gb(VF2zc+uC1QT_)S0Uio5mKTamSK*BjkpR~gQGy=@V(iHV=861p5^dvWLpInjf@VM1Po_)Lu z`DHl1=c8|r;_1k};;W}y# zE(829=!@~g4u}m=+7Y@}1w0{aG`{NgC|yCt&jRqX+R1brQ{;YvZADw9y}2`dY_nHv z?o;=a3@cB{^MuuM1qct8b$qrtMXo5@eH@Lxt&r*O7EMpfm?laHCAIQS6U(gUZl z@Mvhgc%*nznkQ0LE1=F~k>axrDpH&_DfulLplFh@Hs?7qpd2RsXi3QAIZ~sg{MADb z1-_D!ySJqNXJD)Tq!LzlOev&JnajyZblMKJRXr-3;c`i4SR>h6)+~48(Dp9N=9i~v z%g^RQ|H!h!&G_9~Q(%UynZ{M!}w7I@fU#_Fh zP-kJ)4$o)XvRTl!eC9C(3Qc961)$frSg4jnZ#bFlKRGlEkSdNwfB!etBDo^ehqLj8k%uUN# zPE&?jb;@&W!Vvohp@!+qb9&3IAY zLX+{ zL=I349HcuN`^3Q)1qzvg52R|q43Hr)?<(2W}IsndnvMWt#4?$Z$!s8;^6-)pw=5k%{WXFOW> ziU}@PGy=Mt)=?-wv!toAy@v?=B*bVCcq50Xt|b8JU&*Fpex`Xyl)XA7I_ed0vTLmcnkSqCG+g3$zd} z92@-$*%Uu)$O=3F958F#hZAxXGI!B7(064QL~N(n0OPNQVaw>N7L`r39K3}4scfux z(yb>}mMgGW=3>Po_PViBaQ(b*wa&EI@IH>TeUd31P=L}mm2)`|vbp~5@LbjZ*y z0u5n4JMVyydW`dC)Q*iOpxAkmzzS-Fs%s3Khmwdvhs|6CrEtrRwFpjrkw3%kx4-la5J)U!Nca9~WdDhP*qkOK|itlKB zRM4uO=ILe*L~uxoN3w-qtVNQbrL9tYL5l~2H=g==o$Rt<<4H@Ou-UAD>!(XCK7rP? zg2~{V?dRz@#1g)K9PKpEa;YUB5pc;E|CtUGwGI;L?jn)IsU*Ar*)+6O)TFxl`~tpz zX{b>2LYDvNoT1BQeqVa}8`aR$cX8npja@6KaBYRciVo`08^}SxVvOGgNFD1|`Ez}Q zqadp9TABcJ3!P?)5`ugLPuQqHzclD0jx2=?=tC3&Ek@|kZD1sMn&@HFzLq~>r;h5+ z6>Hs7vdJX85d^aneF_;Oo5!npH5{?EPXzMyb#Tk)@*|0L^T!|<#PJAU+mhMGmo7h1 z)<~h=SBLz4>hf=F6cjcmhaV1|U66&E#B&da_^&Wt=J=c&@%$vvU0E6pQ2a8GF5K+f z3+m@8NtK2inFN2rnMWM@J$3zrtfg@P!vR9XpuVz!va;j}lR@;qA{d@ePk@ag_K~)@ z*yfbVgL)~^3^e(p^8pDS-=rk#w3bQVY%SFR<*O5i$PEiv01ikEHFz++!KGSihuL+7 z^LE9hCUCZUv?r68yn|KYo7(wZ)V1#hqdv^s?Kd>&gPUx1;{XxTvhm}|6o5oi1ycaK z&Kf>3;LW1R!D*Uj(#|P^>9imy3M6)=9I;@WnNA1}=#@xlifNnL*LC>z`4DQlsdb9L7?k4BVCrYKgWUC` zqRDk%f=y5|0JX8Z%F+|bP5R?yla43b0K%pUwgE2bctnXei=-^`!bgI1-}niKg6F^u&AE(p92}nCVUKOoxaIFmGhQq%d()0CEQ9iGzE$sEIo@4i_!u+G#)O z)+v7Lr~P0$hy3L-3ewpkZo8XDVm&_P+5J$GQ8va#`(I4ih%!6O#CSqR}v zMPHRA<6X)H9^}(0kQ-Yr*+J;Jf3OMo!)+~k;jU>;2sq|KUAwIJ{>5reU4UWt|lH=h(3mHQf0dDdMHJVtm;i5F=HvJ+KPF zr|rS!dcm~T$}_2&eMA!zl`@1LNF9{jmol4;(TM*%@l1&;z31RnXRo9hK)G_$fj9e2>+rQ4YJnH>&qNDQV> zrCgz zFg^vSYwlve{hcja9redPzQ&*8DaNRdCKgU>%@Kq}y(W>sWBnBOuK}#b*Adfum}Jg@ z32gI9QFNjYV$62UB(I0gNcLM__+Q44cW+Udm?FS98V2EQ*>GC-BZo#fWu^(QJ{a+7 zny|TEz-`@b?x3Phq8UgxQ-536v4>8KZXH64cVyC=li1pk$JnT!`gnhi`{*+tj#km8`=xi6(^Utx_O5l4neg-aYtxp zPvXnSBpGj`wux&#SjJ$+21S|=KCy*Rt$YDvi9! zAxng8+4X2syOboU8!}}O0Kz*5DE&h+fic5EfE8V{2A`I@DW7B}@j}hCKiIPI_H98k ztwHKNg*lU6ywc{@{kLno-{^-NjA48ax9KNT!`f`44UKC~tRltHUDv&E(Oo3))wFUU zek-aj0=(=?_DN?fc(!*;y@X@5g4y;+ZIH|Qsf4pmAj_cyQ}Wt+lJ~>25)7YQk>Sd+ zKcKb2LjB`nYHpdLNrl0}`QS3fcY01U?_OPxKrq})p+(cRJG_mMACjQwLfl|3;IHf@E! zCBnt|(i}nV5s~}Z#(qC$=02N^q5S=5|EqNsv6HnReh~>^v!@p-#m}~%!itsrC3xaX zW@k@E5F2l?n5f7T?9m8wmc1M&LD?vLNucw02x-V5)XXY_2xe}22t36V+It}UuWdv3xyqgUk8#DSCvXXCG5 zS&Shn(!Ss#nQ%QwR*C|!!?o|~KwbFS@HV!mUQFEY@G0eP5nm<`8`Us*FeG)wACgMc z6jT?2EZWJqw$;ZV)=?%hDR1 z{7Q*RD~Jg&l>#t9*%BRc$?c}i*4G}?AGx3bAEYd`I^j&@jsz|KL4dy@TlhfM2igpf&IeI=eI2kUxfk^K zF-)G`W$?Xr1k=}Q53#uz-II7UxYN(e4Q)4{=I&4yPgq^ssIPf6S>S*R<`r3W$nHkQ zl(Ip^{4U4tH<9mwjFlCj;;~krOu{-7PlrTscckdFUh|(lceqrJf`Tj)sI$-^saD{S z_$n(Eotzl29H7iW^JCD6EkO~b@&kwH$%DZ@`PESe7d<|kjG|rn+2Gw)H$I^KGj-cz zj;fQgFS7mBKT7}-4KpEg7Y=(E)e<}#N$dEHrdJ(N(%Kg)<(sMHhtk>ZYnp}b?Gg2Q00 zV?zQU!>lVrI?#GV{aL03;4$;$VPKzJkq6jk2e4T{u`}k}oR3P{5#(Y(c^}Or4IGu; zokn%~)@>|!@_Mk=~fuyE6r7;y=G;z?Au=7f1)5j15Q!8?R3^aM9(n+%H-R-#Xdae#cNAe23`KRLKhcxzMtU{{9Wy zF%>_wrX*FA%}RcL{f1ioF_efKq!%WLotYQtQJAoO$L1zEfw_XVK}i%;i6|Vm@P>>A zCG&wVvx3b^m=(BC@km3!vgPry0z!42+Ax$k0fS&f(Oio_eyL$7-}zgO9szzqc*U9K z`vGNo+1bEfx+GVr?jXsCv#)GnAhAs!ktPU76BhTBvC0(?m7GY|$9uqAL|;E_16 zRK!L(3vrcZWN`2P2rAIlrY_VTT`Q@waJH{y`D3<`L+8k^BXeYlVh^NAUo2cfUtnVb zDOyZWDf~vTWLq5~P2w3X-Wwn$J~~HtN*(9A__UB&w(!J(;Kyy$SA5bt;2toJD-`tV_4P3uNzcDr~PJ(}1x7RG{8!yilL-1W3^Krwo#C5z|g#cJ@R zhX-;DJ&|*&ABXJmi@H!7IXm?jiSRK+sq!q6aMmh&{#uo0pP_!H~cf zb08st6x+;TCSFxkRT2{l1FIp47UUfg!VGQtq)+WYe-Plnc`E=9w2^WXc`A#N^Re0m z=1PN@8o>$$ocFHvD^mUi7APA-o~#Opp$b?Pq%q{rs(=_O=vKJ1sQ{x=KRODzw8ttb7dR>og}4`%m)HDk zmnUCiUPO1op{O)93J~J#CY2&0qf=$IfdjBi&Oy!=rBP;3_P?rD@Mr_T6I?_wn+`4` zIzreY9g}=wUquaDFp~ebQ+b+A2iz8B>bsj-Vdfsui==+8R=~zSnkP>)?%7Tts);n` z%ECdu1phZHA^4xBbC1a3(o#X?*i~s`5EX6hZ{n|yG&?ttT|GbdK$q%-b;?QJv z5z;)q!@Yh}J3@@I$83>B%UJ4zwXHX_Dnoolf@FZ2F1Nqlb~px$DjPjdQPh84^sH6| za$NF=IW85Mo6X-PTSz=H2E=u}-V{`QzryV#Bvc*2H~?Pg%(bLK5VZSbbBT9M+n_bwab)N;+Y9Lt8MqW`a`kV(}!!PY_f***9COsH4|koij0ed)kar%9792yQJZzi zYzbx9CR`{xUh^c_#Ccd#cIgJ(E@7mN#Iie#jcC^cIFaP$g>8 zh@Xg`Fpc>KKeHTJ?iu$bc)xisc>n6fqesE}#?o>@Qwi(XS_VOVsB+~bxC~4T98>zPKUMP#ddXv{!iHEKONr<2i24o^f29%#cO+bDyVqUwx>M0&6|W;?PLb zd+Ic1qW77pMdv0t0k<0y1c9VJgUn#mWqVmT=C{6 zkEvs1s&g?{5zhCG*>qVCI?reuLX-k3GfPh`)Wvi%1s21Z0hKa}dhRX}MvQBd1B}r; zAjK%6PPD=fM}a7$svt_;(|-Y1ofqb&Ub4pUBs))}RzP+>jVGU2Ws!ALN?@<2i%zb9 znE6hoDj3>xAR-uIYN5U^3eu2~ zh4I-4@PyDpntJ~tQgfT$#Om#nE5es;_{2Jk*Vv7@T?CFl}QlB$29Qo|>*xKSW# z(%c0k{5EOb#9}VbnT4kHh8}3G2_@QTZBubDeH?S5_ozaXfmb()S%}pKksJ6V)hD^9 z{()2i2F8$_?aV}c`K}oxeJKKNGo1yY2)x727d{gkE@{o8w>o!NNRltBu)ve+Nl4y4 zxhgrqCsw&p(898vhI*h%)|)Uu2T55oJ`$1GYai}Z#-kjy1R%TWQLRFcl1*# z9>l6>Qs+vm3qs@GLJPTh1N0i$cJ%|rpC-doZW?D(tV_yyRO~tb*#lxm?G8|CmUaZ6 ziLwoZmU@m7a3vAdm8Xc_KQEXzU*AI11(Qd${YK+$0m8D~+}o}1H$O0LpE2q7#LVQf zajUzv-E>jcV&~x+(EW|H$7PlJuFW3*RRI;%zWf)_g`$#}>qdqV5Mbm=RCzLLhPhb4WY@%&9__ z3cq``zyi)ihM*GNHNY;z-|XU{>fK6g$Y!W%DiTsQG#Y;yLh4b{zR@T$$vY0Ltvru! zqINwn8iWJH9LxG4r)J^lRHn9;LxP=ZD^xg^{V%@q1fcU}8pUOg%h3Rik6*ugk%OHN zC$(V0dSlEJ!AO~f5>b{kM81(BN!@17U_`Ty-}uYCRh_@>tVzS-Ml;SwxKovSmh}ma>u!cw-qN$4FvC| zT1Z@FF6pK}`C%v=>z;Ek&PCjnmp}3{oj7XJt&7R>yv*ol5DN&F&kG>1SXH*;3 zrDZh;0sd;Iqxjo2e^JVoTK$dfr3;pKJv2*f)>{jSh1Z0Yt5qz}!?`vMhi%rlaKn06hD{vT{2*`Ev9i;%J;J}I4y3``c!-{W zqcd(n`ZgWh*=m7XIn=x*1xtkY)-O=9r9Nf)LI9lb!*WlcFIDo*w^dLjPXosziq@PfO&YoJcJt4)z0eMA3o%tXg#SS>e-MD07Gi^CNk#+x#IY7+$p*NWN&>* z+0uCJuaFy0kH9A~wOTMkv4Q7YNrb9*?rqLn?Vvk_T)A^9>Ru{ny< z2~?45^OB>ecgOQ<#S}>#kV+E7JfZzQNcg8<1XJrhLIc5&?5vbLw>nHd)~^Q1)JpYZHIBgYY3)Vg=Ih24 zMJ$rQ5M~v(>j%ly#Pby^L0#T785X{{Z(cUZc+x#5Y%<;RSEq=scw8cG6e*VGDWVVI zGT}K0)1qcMX8GGw(Ox+cYHLkN$*IiIU;uchy3=^c*d_y*)Me347PUHA1 zN$2BtaR0O5Yl!_H;S!AcywaWLw#GT|`J)17Q>BF9_IRVu9@ z8YDy9JkEk15E?s9blVOQv3Z4cN-Y_794BVWrdTQy9cxIYzx6V_cc1+F!~4}$zhdo; zjRMV${ZVk!t8WF`Txi8@sjr-6`>>dPve*7dyTe)c%z}gRAqjs2Q8DOqzAp}B`$b+* z0easjmkk0xn(GRJrH$8bj4cR!HaJC+&AX?@6>~c+I`31>rnAZB*iA#y6^9gxbIL5h zbqj?@E?Jxt0FV#thEOOtY9Uw4uaZ-8C993GwIE$KKAs}D{_p31DL$50E0QF9b}E~H zhS<)nLEghB1c33ta-c2Cb|2eqMB$+=q_P`s^l2dPewczSe*)*yzd8;Bpsm72NW3Js z-1LQBMlbr*sJeSxts?Z6tH){<_Y0|5q#AK!5S0xtKZ@}Ruhm9H(vQ!+rbr^XIi9kJ zf>ndIBIfm-=9NL!&!h?-g|bfs%~$YzoCNL8kz*8{3PT(DK%bCvSc`C0F54l=hq-Hs zu0_Z86k+)1<(swCLFzRTH@lBKHZ99<3sn1f=Gog9Q*cIeU0Ylwxl+1QOgB#Ofxp3SJcdek+ zU_(PRhPg$^7so1w_~cU5V@~x+it(ihHFj(s*%-@j3)I@q_v(;lq|Cx9k;53ydvcFY zMH=_`ngPpLo6<%_ZBK365sIDBHFlz_E1*r58iPw<^iLjh zYpvf-2z9FC2&qluhqv02#kC#P`J_761yJN8qr)e3?Wwh&uqlKvVq%g^D0xrvNmzq) zr?r0~cu~c>`92>oI(JM*6Kz%6b*Ctv?(##?@KDv!k%bu_wfpZ5cg3R;PQhoeY zG4Zn-bE5Fe)m}CV{7AGb3aTK{?xMh>MEi21pp=K4qxE+&ygIcHpaFdoZhL6qv5?)> z%B$)^LtBVZ%C0aI2ubOs=HG5it`Wt|HGFQD&^E^iowAYPN19!cv5{hqub&@0f{x3L z0_6FZvriwX88g2JOoKys^c-dJZgZAF(*0C7yud|{nh~~En_T|lZ4uz1qJZQ6~TslD#R}rskYA6mK7Q-<49?DQE-h~ z#5p^pXyf7V78f7m+{}7Pb&`$%o(|7XqHoDfK3-t!^K0Z3;2|Q0muyV&PfLtRxhQ~4 zK|0e{{8_HL^rR$DZsC(F(q=4qL==~cR9^GatsMKE%n1`(Oipb^^22W&LNMPUs9LP? z#1tbGz3{;ot|Y2c2M!1mpoC4;Gfov7ZuWV~57Kcbp+S`znXjp8+>&A9M`d1NQU#TH zSDg678H<+ucBaFC)n&*1zqcDZh){&igx(hBTMC}+7iaO4yL1jn7T1ik(hU*4B$HTF zh}CzapwTkzhaj2~S1x{jHByDzp0e@cM^|3)QUzUk7cV~H$s%Fd8KWGZn8;>Z&XQ!S z51*!f8Gn1Ycw{u2bb1Ob1p)@Os*BFE3|SvXI8jjJY`*|y>Pc$w&uU5dGcK2g(^0%@h~d$;=?R<60Pjkn!Q zCNW(oDep$a`7;5l7$D%Qnoe+A`cPA&$^xa=QC;QZQUtTGgO|~X6`w*ta zxy9-)P*Jn#iip?(`XC-qQXIzKO|9 zq>4*Fz`UU;IVE1?BgA5It$^Ted5+k2*Nwqa%13&$sS%XC4E)H=E9X% z=HCp!%DY|b)cnZByHBnNirw%DNEU4#w_4jKp9oo26nm*lMVAR!VDNCkHI{i!{~7^zgYX(YA-wg1rCVzcWbd7E4`koN^8O_)n$mC_jiFHfbb~|7 z+gfM1WQOE67oer5lY;lO62kk+8HejQ{m8JpPp(R!_{8{%93ak{MwmN~VaavIcsBOQ z9D;rZ1WcPmj?yP8$pF9I(0a^VMMrht(ftdyqdw4i$!)8~gNm%VkTH-y@YVf1;LM3X zxQ+m9(e8iU?bHG`anzk(_B!_?#IAtJgxFsnfbfYHZWf@05M1J+sUzE0h!EX|{S*$Q zWXJ-EWy^m3JQ}D{W2v^D&K>91VG1F{tz^Y^Zdn z&Nd|Moucoiz(dB_eZ*W!rOXj9CfF1xnK%vPA&O4XA%Gb!lKN-ef@y~l)sI4Pr?g;I zGTnF*V<&)SV(hQJI^|Q0eR;D0d6*+$w}=c*)vZ)a?&1-w7Ii6HCle)_d*svY_RjOX zbGOz-Bszm@E%)RBk_O05;$r50Oj$NAo}$G5EZlOnQ5D?Mrv&_R(UN8BqPDYAi(DX2801m+-oOj zTbxK?x!l;plN7!JV?Ev5e)V2w*&}AU(kMVtj0bTFGHUje3ED!VDQlteB?DI*uYzk& zMlr6|DoR|{ks;t##=5xK83l;5wys0U7t&1b{$bIZTEqcxmP?>eW*l5D8k54aCi)9~ zw%tke++`!?DT?^dyJ9xiDgvi0d&FZ`8U=KXxs_pus7vGF)?FVbc%SR`Q9}6~2G;;B zSu4XbQQ9CMYo7=h+%WiDq@Zv}FT52v1`Vqo=$OU0*zKdBDGkJEP%7xS^zU_bcYHOjtH(Yip#X7gj_ozrA-gIGx&Sr!sUb2tIw|2SsgLgfnC z4)M*F=Jyf;A4i1OF(@8CEE^TeZ+!@0a=lbSyx)d5%O0^;W>AltDj67ta6i5rTpthl z6gFBO=c_Bnh!%n%bQbFkh1t zOM}NHat3~Ej_sQ!AN4>TvdzrUn@-4(wz4;qA7=sGOdA!jz;rR>am{^YrJ(UqUxS^y zwmD|;!Z^6tlXZ6fqd|zRX@K zdjGysP39!<>iw}#CxMkByON9|$~m2eu~`Q6W5jL1*KkTjtljOX)^Hpc)Tl#kZ&k{y zn&vFjvmmQzWRN6?1!~|2F7AkT2V*)T5Vx}07@iCONE&={MP_{~9%VIGib#Pe{WH|@ z+btpmt(km|yI1C@p_VdqS4H3_+Ei%5I&a$M5NGsf5zMD?xl> zuWJQ76Q1#IZ({6$vyk3~s+DU-M%+8$SZl|~sgX30;B)WHDCY~;$_IYWh>O}#r>-C> z6fXR2)0sPlb}7S=+QjC-C96{O`F|V|GcxNx_9l~?@w=aXqSe2CGro~#OgMVejt4){ zz~69*{ z9*PPMb*!GQ4Q3Oyx%PB%sAf3%xpey{qHSC)Xsn%05x$eI+I7IR`vitKJz-yd<>I3WfAV6{VJ3E&f#nX+&Io+SJ2lV2YkU0ScN{EN?CW4%C9 zYjz=g#!}phhrPIsas8?wBQXjxsEsgYc5 zUxxU5nN&OC4W&lxozaL+kNE1^QvKyrY z8}a4jmN_d=y8iabuit@K6!>gFHVZDM8FaNE^Ep9|JwXL0o%GKpEV+n)%T8U8rGe(~ z8Urx^iJJ*`NfWhg99_ReCr>Q@#1(<2aW`}Gz z=0a6?fmj+xZQILync|li(5GA-J?~=j@Sc*&G@OVhU48rHiahT=TZGL5eUe$$2XNT^ zcGHg$=rtCgkSHE?5rpX=gFd|ja1*1|Uu@TnHK;3@5vEJ3i?NYF$)GThJkdEYBppZp z5_Sa@aC)ItzB2(ReP0p}g10MU6)wc!3z>GT%`Z!{1c_xUl7jdzV z-?6sBn5%NUT-Bw>n4eCEXNBs^*|3H?r*$=-lJpu6~xGsmXsl0Fm zQinI33fnZ0%IZ>HT15$yMw+E{=_D}Y2uDN3rb z6?v4X-H*z~!Sec#A*<(QgXId4HS8+#$kkX~D!?|*!vXvBg2lS%XP6r6V}37Yt&OB2 z5__&w{Y6)aOtv2zz3&Abt-CPF*EO>f-LTAJn-CPMGKgyOIsJh9ryNU^XV{aMfPHdB zvV%`Afz=|d@Hr8uJ&KDuW))@}Gt*=M)pqMF!~m)qCbJ0?LF?&oG3A$Nn8;wveg@s6 zo3PGp7f!+W3~bEV-a1K!lb|e&ffazSq>-5nyn6k*e10q^fqS)2sD zPK4DEmrBb>7(3x5UyZH<63pHx3Ojh(^!O@3ax5O90Q%zPs3%(jQcY${@cOaj5gA=w zF2KZ6Rw+8D#j;jmTeYej#kWKY`iAY&qUd2vd|@eims7x z`AR3!iaHYS+3irapb;BR#_b-x@~1u*)wmzr;$hc^+7t0T<19z;1G;F?sDv3oDH8M! zMI^oq`eoZ1U%mxPsLWJ=++l@`Pv3&oB5iTIjs^l#YA15S2`*7&ny?LHBYKTynbjci zWTt2t6ATy*Mj;|$fvM5B+GhUO^Qo#iM7&0U8s1U(5+C%?eb(&aahqt#I^b=jij;zy zeaMN_((FTQ1tZLPPTC)s^Qlm7CY`YMvCry7$;#N5rvU*~1y2JPP#%HW)go2t3u|OR zHUK<~2tNorOp(J>Py$IY=pOm__*p0lv1WVy$WMwcNMIaL>Wbx2_N<>`Oi+4o@BWi} z#7kFK3cAvB#Jwci1%Rnuquw-Uv^m$hJ##p=Mm4hpaK33d93xj?@NsT!ULY3DYT5lX zo()fu|MtmM={$U5qbmiANV~y^y`bH1+}a;U)B_X8(Mv|{uzo*m--3(MnOR0SA9V+T z3d2yDDKSV8R9ksms0@BihDeYZsHq|R3G;Y~hDjs}IWZf#@UX(d5!biuGwVt9pGc{K z>c2}CK7q`YfygaBbP7o zg<^B6IqP{yI-(hQtZY1yoJ-iRePP$O_xJy94ahcbny}F+W%;I$Qqq&H0Fjm13cP+x z;u9P#lCVAi81KX&8OPBLokBGEPQY}NKBu4Vy&ev)`iZeOP#Kl$gt;31=X^DA5t-6F zs)bn3P$yCVMl>tg369%jkp1w6 zUAYOe9m#ieUus9mb=Z$@2^zhDr6+odt8hBjST4;k!DP~>NG%yi4L|E{1X5$ABJ(z% z1_`T00_G07X#JLi;rH+zons=rJq0u}j`?ifqWa|t5c@$GqLNDmb9Coo2l-l5Y$Zt7 zf+EDX1Hg=vX!rtV37p`HFCeuNW(k&Le3~V!t`-nPc$#&oP_}%$LBfM=Hk-=sPe%e2 zC{%KViTJ{upmzQ}xU^E_A_ly=c*DMucz=vr$|iwTKZ|IDQ08Lr>LrCw7lYNJWr2@0 z9N2rGuu2bpkj_^1th*bfW!uuu5F;E!^*Py%MM8QmCUyerSdg*=tC*tcGrkEy!EAl4 z%;+6Q0Z-GDnODK9699dB6`)!5xT->PSlb2c9!OJxIoj?GhAH(sp^CeMn}|UM*TWWm zWp0wT7v9y!QsS6}P44{D?J}d0Z-pZWIAjBeM1YRsjBDk9MfsA&fhV5=^1@~Xmd)G+ z;1gtAD{#yC=oniLnudtM)~Ny1sX%1RbXDc0GKDF@wE!i5m=Ri0U-kxDNxQAt8G9## zLC~Rx`A({cyVy1l`TjW{tpf8r=BMLw?gUG3pX5&PEP@7~ToFL5;S&>GD<}~Fv)D7& z{6aRl^cJAdZ5oHMc^o?EOFwlAw&{=4;YYrFpep2@hiRdBKifJJdNJzIbV#&VUKO&# z(R<6KjVGBKJTE936;SbVX~QQtiRpK^s3! z3C1;tori>e7zok;Pe|P}cS--&Iy_(u17G(4`29S-A}#E2V@!h)FO$(Ax#mK-3@7pJ z41@UN_m%@8#KsUAi?CD;Uj$1fK7A2Zi;Oi~C1OTwPFGYIMhbpE zChca_K54-bCQv&yvRw>{FU~c?I5jz?3@A2JP3unx5aa-B2NpM9fJH$+3>>-Om-zwt z$Elvb04{>Zfv1b867C5udVI1%i`)}t8HLTsnQYUC*qnUOjAgFjNI>@W+|1vqm0Tt{ z!X!@F!g+s5SrmacyH7;<$fKqLMq($!Qu4Z4^<`%KQroRtB+c9N`zA9ZYl>HP<0B0Z&?}d1hA=pr^*8xQKm0+r1so)ZfN8oj>Xa^Tg zkX9VQ-zDBv>jV@2n$$W@x=E00hP$)j#@xosQ?bk|MxQ0>7H5RWM8+EM|*V?SYvZ`ZsnM}vMjU04C9 zsle9`Jkhv|68$6RPGqo@pj8!_w!Of#?MY4sFS9n}(@mi53_Q=*j z$j7sO?>v=Q%f?@Gtz+&Gnmum)+Tj>A8xnPRb;UMvc1gEx!wL5~M2g^hcxfx6^pG7{ z+KYEHV0E=*s`g|Vz};O1%K%H&D;}5D*Om&16e;%jDUML(JxyR)Q`x=KB4!lV8Cj~p zL*ZKl?z1o)(jA(nQZ6>$jvpcgGjUrcLTfBzq2t*s2ps;Bre7$8WhU@+i< z#3%Gvq@Kch(4`Ud%(ZhbF*42^OUIRo9m;&S+x(|mOC%b`q|m;J03hcsx7T(oXA_Oo zuX|D7?kF3gBitUa(TYZjlnKIb%+~=N`u&mxi!Xx#;$)=)=B_SId?J!XK58V18?7-b zBb-R550H{8}}1K zj`VbA@*{jPH=N3y*+DRr+mIgYoshM3DJ%jDi1L8)FZteVdABkrGZ`hmPgnyp(oCsm zy5Y@Az&`o)p#h{Hy;=#ZEf>@W5LN&K$j){0JrXgd|GXW`S3r9kOb(|hLXd45=Yj$b zZk;iLTqM8<?Bp^!Xs6%P%a>4>)`pz#!e{5 zTb38ApB|Lo)tfYIK;}ZmGjM8bE>Tt+g+uh@9$=qbk?`=@Tomn%ogikKfA{ z5uRKHINP-ftb|z+;j=GTFPJ&OD;ZAeZdn*R2yKv0ruD$Ivd-$;A3oSUaT>y9VJJxf zP>51#&+At)R1PAHBnIRJ_R+7eCG31gnR$YefNc&ohj(fy_wcu()93}g%iw$Mh`U1~ z^875)G>dCx#%2HqHG^Ml=!Yr=xMwMs*WrR1)@q96BL& zM(gkY#_Uo^)lPEB)Z4UA4##S4q26ruS%xkdFDw;%8Fn}1Nl=kH#V7fpXgw&u^I-yr zj)#jyX52xA7ui%AnS}^xD!rsE?DGKUJ8BD+qp97Y5&Kh$w)PIC#deXH{iU-+q(;r! z8bpF`#NvMchN^XXNgbr^G!eVCAX^Th;D9~Z4cI5U(0R>SilrK#G>Ib9BgPm4Kjw8U z@@Bf0h}@#Hq@X3V8DG_*!PJ{0Oekcf>C9HOC?K`z6VeK3XOrnl%I?iIL&CCti;^tV zM9On$7O@ursH7oi?j~x6b^>}N-Ck&X5XqbbUcChM=_IgTM3pL5Cs5RUio&~_nEk-o9NShnhuiJ=)10DIZ3RdNTq5y#lq>&hwy7f_; zeQib1Xb~+LnpzU`75Wn`k}j5>FnhQpS~tmv4)zzs*K3ErLBv6WGq^>${Uq+|W29y9 zaiX6du)7g;b>Y0{_l|-D_0Mu4<4f|taWDD*`pLs1AiBO(z|_x%Y&+C}+3XXd?0ouJ=%CvVaawqNyk>i)`UzC%A?0M=P>TOQTWx z0|D=sNYREHfQ86Tzc0ydM*z0?c268RpuWk+k5TU5B>4pRb^q5<5 zSvK=~atR-~R`!-?H)a6fc1nx(PfYjO#`8$yr9{7I*p}k1cVh z9?2NrbKoivO?Jki5eV-KCQ6+A=>t70q_??W;L=`%nqsOkJc_Zcg_8j{qLoNJx$OCP zQsn1(uYw}Kd)__bhY7GG6r#%1s-vIs(zJg zxrhARJ)*mpvF-3%gJ5&-ki35yL(wi72)#3MTz`IOh_?=KP3{vE@r}bp)4|t68?|DgO|zQKBey0 z*M6h}^#LM#X5X|sY{!_z)^S)Z}utthYm{KVjc-UF1`uyqgC~0 z%oaEd#-=Z<^o0j!C#C{U=36%0{-B%Mp~z()Yfn1(Jk~2q6{+)m>fo;z$)lfz6h}kb zlE`v*>ZxLA&l%7TGa)LmYv+1zTAew__Dy{20(*r|t5_l#x-;vqDfEj`FTygh+ zX(R6k32b>?;UgBV%!@d8@w;N&B&CKD=jp_?55c&NTQw3t>2CFB;2iev2|20xNNSl0 zg@K5tc!36Oey{8};Yrf^FLV$11m=q5i4Spo);=T;nFQ{;w*E%Llos{-k@22!_nX!( za6aIfG@T9JAhT#wr(N(;+Tq#9gA4mEdH%w2o>|M&rFCFH1uVRDAu*MFtUW2+lj5tO zbnhN(pD16EGI2@)P)t~gXNMLw;-MU(Ps3Qk)t`IV?xIWV%>8mtE@93A;E~SC>IdEgzX!W%&*Fw?-ASEcw0~`x?M=c!{a5o zEDTgLCd5XkwiS1#*6icWm+nvA0_Rg5<^qMmu++<))+{ez69|>hB1DdkD7W#4a)>Ge z)rz)8)e?D4B7xj5A}f*q$zu40;*nrfNA%D6T;q$BjH$+&r`iF<{i?*2M_h2DNW_al zAXfoV0W!2ZoX|r|xy-GcP!n-)AnMSi%qaAcFnW)C9}P+b8Z2?Ixw2AogzX+RfU3!p zkcqMwvBs2#QUo^0EGP&rYdbwCxx~;}z#eeF_R9!`IWVV^k>g1?pKP>}avNWrpw{pR zJ#H)&WTTx3cW`ZCMdlki6A9^dM`7BAw(xC&oT@-N8pd~IB{Gqiu}q8~#gZ>@4Q2YL zltWmMY}ll5cb1xvZ+tsl!o1*1G)ZUI7-1~IYc4kyHM=Z$ZIPqWu>`58Z0mx|{g2bn z3zeT9(ZY?T<;T&&`!(2fCe$1P$wMx9pifRqnHD0s8-&Ms4Df?B=ckZAC-B&rI_dz9 zI47Qc9h}>vJqx}D6RO&4-a_pFBvD4eCR#N-Sf~_LGIqw zq#n`4jUq`HTbdr=HaFSH z)Rj^;96ZU`6Aqb-{j1*rK4t7T8U=U2UefQ^t_-QJl0^t*jF-l8iZOvr3d&QFJBw~>6Q!1Lr*K?U9}Z=Wjg8-;2c@pvj~sQhY% zW6E%f_u3%oPonxQyy5iA_)=EfwY8m^ts_7giBcMA?(T*!%)=R6UjF^xxVcezqdvp} z3j_BF=2&!k0V9eqK`#ywZ{ESbd_~E33usv~wR@6yCqSwo@$Le|BjUJGB(`xl%w*d3 zn;+l}Xv;vIz8bi*kW}wRO}j!2tfzWF7olq7?A4NyurHFYwC+yjdk3w@tc{ENY@`S> zDjNfyRNaYzDyX`<81M-)t`una9Z`89x`))V2FgYeo4?iX0U>3Qd6GF#|Kk_MMG|s# zr+70sbXx0wY)>foB-l5&kuW6F5A?gf4+H0{_k1H}j?k0RyM3}l3tzw9_lXv+6tD<) zzlr+a4Gr;`-+oR*#8FDH+!gmo&b!TO$@w#oZ^;NI{#i55@*y9tf!=Xl;(yOQf$l3p-pT$D16jg zmJN@m2*bZn3+WT~TPt9g$Vt5}x1agV)144^QXcO7GYMp;DWJ`D253c3dI8|?C{#9w zD#{q~4dtP@;;a{e_|{I|h#t=@|9CX>yi?EA-(S7f^Qr#6 zv0jjwPv8sHThXBLj?MEq$bW_uK~!>byYLLbii9GmQEW{S*bOIi`KG&?dJ#T4Fvk8x zT1#B?!SJibMa=>8A4Ygw-YtErwik8!9+=<2Ra` zJgqanVG|kNn2OBe0jw#LawXTNYDu$xA$NJmNJ4HvJ?@HMX$ZJ2bYZ(0;V`%*ao{@; zKsLPP9mjzv2bvp{9T@u3cW+cf-`(1P6_3QmjUs>S`30pp;Tdwr{ZwFP8s=`42Gw30 z?JTxc77BjB{nmn}(2{nN_El^6$sE2E-3g0K(f#!=7oSMuM!|a00U(dKAtpv}r125$ zUz_m(ya6ae@K&YO3I3=t;DeTrU>S9zm|bUmL6Zq~gj|Ky`>heY0%}TlkFo(9O62z( zS`upzm`g%R0)^(gY_a{%e+}E$fB#on1UcO@Y${oM`jT`fP^uv5?(&CEII_qn-3_ab zaS-8@SzqMwSG(5ickztHL)5OA`%$A?Y`))d%16)pAan}PA+6vG?lC?|sQp=0D4$#r z#Bs$Z;#j1E#x1cGjXIVN?uU}^YA{{l8PSr09J3?^eLx5!ZN-hYmDAMm&_+($IQe+c zdIOtI@X{zbFA~YPx^km~Tr|d%aYy{5!GO+%7WP^qCzwQ#eU9ROJb{VNmlF0J@sID* z#_C@LZ8A8!EiQv|xZ;21kv1g8{I4EHQ9tUGtXP)43C`OmSD@QnuCM-!N7Jj#w*^$V z&lm^ydrtCALU!R4J$4Fr-w7Yq!CHd4G8)WJ6d5AfQuqGc5 zDQ3t*GC#)sI>zKkaYOTT$3LY`ny_i2@yX?;bajTzN9vQUOOH`&zxE0GuJ-DAV?zJ^ zu}^Va>MG;-J!#e3Cs!n}Jkl^W|J?ZUBeIT+Eez~-cbCoIQ(YaGzbv7pL@E;0b%+WV z_uukPz10qe3pH_zNeSTDmoS>#5+bC+_c2o%IqB}iqakQIWl)pIgk5zUB$uImt7KQr zX?hZ^w@yRx9`+wmj`4$@pVh8L~@KzQK~NwitC&_wlJGTRM>3m()RKG zx8um4&Zcc)O-5Sk6fjF$9&7G4YK<GP&;-#my6^A{wF-*EB zAI{JPW!a`2RL;?O5}LP9uE?|Nv0K?J;=^t~cplHQjqFRb!7<3BFr{;)lFWz|57_mp zU(FBcUQ|tRYU-QV>_`xphx|uON}P^t^~SADXp_RtxTqfzNS1+VMITdnC=HaOPIB5$ z5t@IQN8Drcu~{UVbbbdTMN9O9T9c}d|N1{Cf+Si=7rn!o(n9*P88FU&RzS}oEV_6Q zwHNN|FU-XB-yX&!c~^E4Bxw93*xuGR^wa}*whWjMQwk&$!aYB52*$4o-d4PFjmlZ1 zolQC){_2M&?LCah540rK(`?xuHdST#RX?SW1R4=ncYF;ek2PTWntsT?r^bZeI|#xa z|I5&Ok-z%;*o1`7hx#LLJ)k7P8J2qD^h%tfKQ*DiG8KzClIa6HEAXK`05X&zn*q9 z6v4o6L)sVp-61H&$4L4ta|wYPske!Sy(yYfxTX+BY#zs0eMWRlic7ko4j0_}H-G%j z4$>uGmYD`~*q(&x?UO5V#d#!bY_1g0|KSDGv4Y-mn_)c>%!WQ>z2@MOO zQY558I34y~1M>#P^YoAi5m?LX70FK? zIVPJc1;C4?9bIg=7idMa&5by^zUs$!j#cOg2y^3P6-)(GWOMezYU>WDt|iK+onf~=h-EBybMYQBK3B}eN?G2cG9B0=L*F@JNVfSJ_P z5V#xG$7CXejGeUw{ng35*7e|qI!030TV$Nx07%76ru)hYSInTSlv6w*t2BxrB=ng< zkC9GF)l$Qo9}TXdD|Bz7zKfj$#Pz%NE^)=$yYqpwMfaa0+gUy!sEA5NR4ho~N*7Q^ zl@q!o3_R)W+b37#y7Q^GzqwLCemaOcC|v{3#*-WBn+YV=S)s41E!AQE9Q3b)OOXA^ z)NMdgP`PQQx{{DwdNILrT{dEk->=3z(?vJVM8%U0;aa47%QjKpMJp~EQEqrwzFr|!BZE;Cs!q( z_=I;?3z!dVR;5zQ`@zVTX6H&8hZ45JJJmsOV(F%fG#99D+;$Zd7-W=%=Gd(wqgpX+ z%Y!>lQqg}_q01*%B?NpT$3?cZt;v*Hg?nK*min?#zHQcJZbE7nXf0!-QZ|Q?I5C<# zHdEpn0KmMQQSNg03>=3D-8V$Nn!V&Jowh+tU`^a&V#Bj~a~;-duUUzU+TRmxwVfO8QwyU97W&QI=-e-guZ zqTa<$?9n*{Wyf)}(>zOM;1PvSIt#^#9MaPwl>%g%Xy4wFOQzIvlngss)rZpEmoX(~ zrjwW@aK7h;7)_?tV{WNbH+nA_UV|vwR!*}Yjl%Cxdx46|R=tS&`?{qvyMR~a=UFx5 z6Zu{*5cUwG3ifT(Ry~V~Cn?I(l)?pMW)R~PanoBM=bh8a$ZtDLE~!_TP%`KWVdv6Z ztGBKlz?$jAw}w4m#L#&!fcvCQ+U~(4E?L?(o}%^ty!&ds3NEG=3_g+L^@5e)St)C+ z4Ci!$_^8@xl8Gt69D}le`<}$e2&6qT7n-^w3Rjz(+zU+Q?x*AMJM#}qE_O57y4G3T zLk{XRk#OY>kPJQfxb^}6QdSbal}3hf2Jn?MD&aX`apDsiUN2~sk$IBZj|$KkL+XQe z`d}#Z;+T6zIfR`}v?4Ho;)19pyCsERcEs9BVE}AGubVemUhwS&W8`^ z-Qrn(LWj!4M~`c7l-$CoJqy|FJIvYG=@xQ=#hmlz|dYDk;XrMmk=m>J@^Lw z@%xcT-4iil|5n>aNDx(_KYnla#g(9lIMrnNkKaE9UpxGc=(-8N!v!&NE8kznaeu;x zV3A!m5qPpHSgwdwfvyFgRt1|I1rU9@^2N3_$sa=+dD$-*%$!g{NE$}{ss=OwSiF}x zo_Gw?+IiBS2G{Xz*)#TJLBJWWZd3%)@M%G?S+vjd{=?ZjAW87?6x>P={VeKoj`Io+ zDcmu*fhVmF)l7?Lv%lQ^@_{~f$e09O(J=(5WGK3Sgro>$rXic5z~{0r=*q`lf^l0x zJk`Wb1!dV#S@#8i;$={kQ1J(V!VN3fEB?q7AFmq#ZYh7NR=kP9$muO01E?y=c1kTW znp_P)W4z;$q`_S>lMQ?yhkXi{5SZakGhT$uVPEm_W#vP8wmmr<*e6$|-0%m45())3 zgX7lekQP{t3WQ~JKb%|P<1m$Y&9Kj;^TF-W5qB=dshIPA4LE%eD&Q&d0%?Swr@O?n zu7c1TflOg!oCI{3&Hh+L?3M1H-;_o^bR(@0eu}>0yYVR)vtq%zL>#2}=?sk$uv8Hb+F;nzRw~WmGp*mJNT`E{=nL6s_4p1!{9G+|t2#(AK;nnfZ zhCjTM;3$~h&Nn{xwXm%~Fb22Jz1;O4h{Gd|%|ajqas6By$0F}uBr-oxL4vH_@cJcc z6PyP;Ay_j{gI5)_upHqJ=_Cvaz!I4lKo)EF-qlh+gSw!P4io&YAkX`?t$bw60w|Xw zboBMlB!;7e;ab)K)5zu$19ylzS@=Vx8`&qB6+8$3vwRhAL|*0}xMQZ4;TCxx%CAVDA}7 zeiGbd*myh;Nf}!<3!q%IoP-_3G$Y=@TqPPXll@z57j(7uIr<9ahv@h<8afn3*va`Q zyrqoZK_w?Nf&F4^mfxmHmopFlv&#v1bM~I>7VMKN5>+1U7Ko~X)_Hcx0ME9<8=_!B zSfxvbgxf(#==7)ckL`A!UZpe3I%Jt&08D9E6p0J-epRltGu>UIb8_`0rcumQE0CPna;_=ne8XYDsF+Dr;0Z`?=nRanf+Jp z9u7zCF(?Qadc@c#mVgC)Ysn5|^(kfv&nhnXsDvnajqN-a^Z4Srap^79! zHy%hiI{C((aE#oVRjdnH2Ie`C#;C0{a*<4&FgHCg65Zv-?l>*{INLujk5t0L!NS2O zrdq@Z1V@KfI#XeVgyWn(;BYchl^;*0KJvf*%m1a3e-7ecTn|;N>x8vz=So`cZu!bS z`*G*KCS4B1AebG&6|r*0pDe>b*-<`X=d)zAcyfE-smk0QULE&Z_6drv7i@DLAD%U7 zOhB(qG{7D68L7iGMQCA7LhF~|HTI2YQGj-O_o z!DT{k!WXVCaeM8E!@Ku^pI_49zQ8Tx$@zgm*{p(3tfdv70BMmL@WGA&A8w^-L1j;+ zh*RoIg^b*aFoQc)vk&+ox%xx<;}$Ih$i5h&2oy-9pj^IqlB>fD^%6dD%8i1@w)OG& z*onAdm78W2$$6` zPq2s5Cm^|5fHKW`2EzkF%`AjC0r*_IYH#<}c)Gb~j;*oTq?9$HbY{vBH9K|n0>3E= zzG#5)J{k;EJn)1Ht}JcdwETAMDRv3ZD*pK7sx*;4VaLq^fQSN3=56%(|LH#g3_0vM z=mQBc&KO51DvM{Kx5UWU?6KR^7D1bn{*2-j$rrL65^}>(8V)Ako(r9GQZ!Z^-)S=y z4m7Fb+95M>nBQ%Hh6J*XX63NBF!)8+eS*l0NPD0U!7WSJWiyEuII%-7-7XPh2=9hR5g7OeK1#{DVo-p;uduXZF5sPpF?cRtZ0%Hga2AsJl4V>k|Y zaPOt;auLCu&;|T;I~-pkjUU|C7g9TLw@cYr@MPdXENoTb z^d5Z@mY0fLm7AHqpL9pKV%fvcwQP-*j96wzl6(5pGk0hGHhr|d&i1LMoIxvK^~TKx z;M($$>~3ZyA9P6yM7G5b-msciw#@fsmcWKiW|r{!8QUYIw^TItvqw|TOMP?M4k+D> zYR$qO%I6j%?x)|}d{00h2#r)>$?TdZ?1gvraW7`PJ3k8GhfI{sg28~F1uzH1S3#)p zfaU=c{?8B`sY*_hbXJJ|=~NAcYbG^6G$03Rb`pHIpa{Ys?LL3Y%9aJSPF{#D#Z z@=du)&zE-sd7~2E30Bm5#O#)q3+UActX|dubK^eGF%P>#cMpd*kfQCuwu#UoJs+8Z z*&=2C6)tXyg}pHBuwZp<*#k{Ip}xP|jI(PgcL@nZFg}1V1Zxez45$MgnLtpIG4WHN zVHY=O28uKN(3jDhUh(FLV4qw8+R0{p-J>JI^3rm_fbrvyQUt4K9{hn21b6E!IswiG zkv1P7n?G78vJ)zMAM~<VkiSnWxWymP$STkGg68a=VrNkP z4$$HFlbzyl3Ropp8)U^{x_fn+NQus=kb0%@ zu=Q!{Y-jIJ;2k4?N#j)4^Kc-483hE2gr-PVGV4FZMc{eQ{YDkc({%1W>;Kh)De(iY z*G!M}G{&=go>fR^V2yzM(QdTDZ3Tk8?S4_$kcWI>AequGLhj)ozf;8l`Pm#mFPI{m zEPRyw@q5Xv;L8tSMNR`1sR=xC2$oh0y2GX=1%&o#?b7{5N7sX__NmTqjd6r{3D&! zHEqMo^dBB62uo{!T3g7M3<0+XBWD^$E(mk$B$u!(1^Jw=qz^H5eqHJu0cFxrJMIqc zJ^37?J$_-03B88_2+Au*0FYco0>2&hO)GLhJwjpPRFyvmQ(9Jgf7!tCB==9?R6_1w zmy<`T!cvh4YPJrcjucU^%=FBvk%^{{5(mNVfy>GXK~S&^O)odsoeEz1-b-A1AzIR$ zqAaA_$zf1TKGLr=pc1VKX&Kn@wJ&eTQe{KLlej-2vYN_Ky?Wj1QQUudsYr$L9-2(I zam8MlXE#59G(Lr+F_06H^UDx8lNS;Mc_6i$3!U>Is$Y;YSj#0`a6qDk-ABp{FT9es zjVCpJB4T~1BDal4HGU$Z0Pfmuo^Kz#6H+bV{S@z#EF@P*B+JycD^xQ?%x)SNG`J91 zT6u!7gP6*?Ta-dH$KK((;r0nt;JWW0D7E}b(0s9htNW&|Dp}5?Wsj}aq?osj#mkSh zIDb$2{r1UKi7cN$Y7wQ@-A4wHLV2-k$ixsKHfvh>t%erWAegxw$$j$-IfwcZbCC{Z zUcPQQ*Nvk~D8cE7U_psSn|M|gm&`7{wE0)w&b0Yo9Vc7z36pLVOgd&&2plOcfz%q! zpb_0PXQ5`R_M_Cm*DX*LirQTyC|ola@6h;g*RVHL`)C^LVOLnW`F~s>D>OlSy>@`| z!|9z=sQX0C`7uu7qRp!S^ohg4W0K?`+Kcm5+lZngJm*GYQGSY z`gMM$_YymmNCP~~hZUZ=cuIS0Rzhq`%On(y!|U*UGaprZZ9j!8uI}l=2?A(i zxa@WI6cPWIK}eQ-!myhK!?#1ZLZ=&)Ow9)@*vviZpYilSYGF=f>d)4tbW&hO4%x8c z!l1SEL|n&Pu622mivA0ohDX?wO6GB<`!9PR|M+jxx~56IipasJat*>etOje7RH5PH zU>NCcla{&A(FJkB+~Wf?YS{fwzo2b>H3~h{jCrB*&w2tqBMygcQ7|QVWQ5jBp?$h&3<9dx)T6v4vkzv1Q0J6!ZVa-O71N^-h&xZi%NE!y>0EiPfN(d9hB1fvSk9%2M zJLjUIt-a4tzmvb1cPAROeh%5fz{3_ljDT`-^qp%OTa4Q28@8%$f)00of>Ru28(KWn zd|T1@&J5#FkqV&h--K8ue5VB53o+BJeZ1-uOeg5a3#*KLm<*c_{o?kt&71V=5X7__CiEfO4a}9jOic^|7lI|Q;DlgN zWLVKl;*r)z_OlB>5|uK614GF;CF>VY>imRHCDi#fm2G&0KpQIsKmyQ|(+&crvMn_D z-p(*6ivr8n6a@^(E85w1$U`JZUgr99S@R=!MWj?hf!`v!b|coiBvGN{J~B}~=2IsMj-Q2( z{Hp!-ndvq42?&{$s9Jm%$5iF2?jNZw>n{3&)QK^I_-NuxKkD*U$k2jSSfX*SwxO<; z>wA0&_7hp_DQ^4vZL?3@w#ei7EV+!X$MiU|&$rsS za)zM#27KZ|KxB@7ZHr}Rgbx3I=H8{bk!;DOiAP58*K(GJ=Gm$cx-PTBR4K?V}NIC-z)RM_sGL0&A|H%0z`j+nq00bG8xD4(? zSX*b2Ac(4y+rzi}^5tt08{g=doleIYFr90Q#x?Of-oa};(_!-Z>F+vn2lI_O)@(aVn(VX{@9OM>gTul6| z^Ec9zW8coTm0c%;BMk%NaZr_;c?B-^uY8|O;VE|w5!2a_!G9l`t{jf{Fl z!C5*uW|7x_rsvrs0*T|#<7b3c(Jg~0Q?WT=*Mmt0>{(4Noa%@>v6C)a&jUv8;0yhf z^2=G7Kk~HJqf4qP{l&ZLEHCmeA|$Ld9xeN@&LUv{Oy}Vd7=+yM_<2l935-C($I(E? zZMF5{VmU$eTr^9YrW@~S;f$LYj5afD^>2@X0b}nLpdme(F3i^WdVFUcy_z*@<{`*p zHgK>|FNX{$Q`_3fl{-m;%Po9&J@sdaUZK;UR=JqR*1REP|KOnG8 zVRWOrZz=!|LfDNXMR~xX zy@#Y_zC*^YEKdmqc~4c;d}z~K73C4CV^#0*8v#eIF5ii(86(2;DI%%DYx zfw#8XE%@XD*hG7H%JC90rR8ygbKDbdcKA-C%+muG9@KW&3s;xUZ9+~7Duu|89T()pRm)+O% znJd1(IN<41@qMX?M_9jQ-;9bFvHwi|F+Y1XxQ-^Y^A$N!7E2z5%w-bca0wIkP-qC# zw1C(T61I!72^>-|Jg-)oIc1M8Jee+!>m91|B=kXwHyzmw^`i8XNk#cS8M${~5i}P< z^2+c&h2)otsE7uC)NexUvL*eoi%4V@IE#OXRIIM+OmgBOJOPJ=I=m$NjR&=(iBntj zSRNnyk0`sCvWy$zY3BP`QZg8;tC2pm!XM^`w5^VcJ7$N>E44i#mls#A^?z-<7R{6 zhVN3!hdrs+69BePu?b<`eN{)olDF|Ew(A`-5Q3zAeyObg)bi_@)aswqs_@Bkg0TCP z$zLjBx1PS=3zMpi>1qT=B8+ZWg&RVmX^e=fy20R&oSK-~R-B~2FJ2rn^;MNg1W8xsU*)L~ZLr5(7+Qgm@b$m-7KB2(nf+}_&9!R4XT*e$p`>c#ku`d;LA}?1#e{l-hvQKDorC|N`oZYG#5U~o9BU#d5 zRRByC#!%v0T`g%n(C%=tk@P!M8GBLEYpJxLu8Ct_I0A_X`0Y(xwulEIl=l8J<^)7i z4Nc6DMGAsN?=KY~^IBcO@M_@?f7(8ItH<-7c!W$LFME;(xlA5YIUcz` z{OKPqd+6r{m3A+@`o~)q+j{+a+_NUX+4N(W;m$|891CFb3Y&^kM$Vzr`Ncd4b|H0nVvE$rs3M2hYYeGe8hb8L_qCd6XpqB*cOo6mv#M#1cK#GrxkIqAIm-cI8v>h*ZuaqIy0QZ=FK zQrIcG*T!B1A-63Fz0Na@u-n5kR8epPz9nv=Azx$;yt^bg^V_d>TU6_*V9~NCwdiNQ zzB1Nbf64isg^17bdXZssjmPWiD(oo)vTCuB(v^Uj*VEorfQOqdg<5~uZJ-I!m7%w} z!B?Hv&tvA^&Zv6}8RV$JBb|L4H4p=TK&JG#^v};(nHM;aEs0t0$zB(NM*skxQ z7D}ZoHE8trppC`i0Z?RH#U)9GTY>Cj8HE@cQg$AsouF6QWHO^$;hj^d`4kaif=p2NmZdk;6*x|s zE9rflq*Ci#My*uAq||hj0KASAcTa%$8Wbh^nBVasLZ1c+Rc16P2}+a#7g^j~i3WmC z2jDqFPnzxa$#YVJ`qXS^ujujU05I<*{YG*-9MJ;C;Dv)f!tfUTGf`2hP(?dQivscv zMnNpyG|u&&9U%<|pco3cBrHfu{BBPq1)4pL~t%7<5#?O~kUG0+T`n}H$57^su^PORkq`jSI2i&VJ1crM zoy)dxu!N{MMWf?PbMKHlC)L1bDs;A`zARD^w~Vrfs+tzFoKLUv8%nhdvRAfs`TbuB z$dXybmsb1ASgZZT>tCN*?aK=V6P$PT9UI{|C-hOrbkE>i4bV=LFc^bmlbG6liokENj1u4kuE;UjMt9R(V`SK+?ppTOc0EbV{RZd_N;7rKR7%+& z8JN+mH-xu4Q6tHb5Ef`#!BH;9J>%Fs>89H!&&xgI5nNp^+G(jt#Mjzr8s=I|-4PAi z9ao>GF&hzw!6o;@(b;bGsCKG~gH`-a-^ImN9R0$wLp}NyRu*2ro{o&SqCnG*^Cy8& zkD5*UeiK-A;Hp;{P}RV4#8`V(IetawLqFAKVYqY#nMG6mNh9IKmCRD}#ld)=V(!a} z1;wlp5UET6=FE)jz4W`k$mZ{%On83ROl1rP{#sEpT?j2HxI)CT1OTn^1-Q8Qsv?$s z*{fu2>BsEn=RryrJVK?*iv{c+11fM_1Sm+k?fDZ6C~Ex4QqdGJw-R&YVEbNh$n>*{ z@>oU!+*WhR%?_~#xVo*qq?<6N2w+O46++3q91BpD`xD2f^?NdZ*<6(XS0;kf{8076k{%i_~cz3$9m$jyX(V=$Okbdg11t@nc;;nW<`rb?= z(QXPqCC}cM7W>NUxzJ*F)t*P(bGeAWvqSraC5i*Qp~tE65s${O#{rz zL7xmnrjDJhSJcyQDhIEA`*klx)Z^+NLlo0MXnxlLt>j*U!A=KYGhjNJ4NEuqEu0;` z|6gSj1X~zC&f`zI(D~#!sQ`WIsV}b;(1*7UPtNz3w(%4c^N%W@uF%L95BkCVlyT-J zvu~-S>p~dzIqsvN88pcUbz%+q{2j)kz*FVcQh9Pc>8IN#yW@WG`@pAu`f?GQVyD|k z%$%ie^45@lt8yR~2{_k|aK-(pwhY`1t;@&@i=HuvAnYlGA5NcWi%t(h5y?5#0L|Z?biIN9a>5!SLeu49KJg)t8F0T@cbXiwRs?rI{AUq5t^ad7l2kr{4K; zk&oYrL?l)2Wkq6s2$S{UOo}B)+Gh_u2Qk!PWPM%Q1s*V=5h}=EM8R}AQ2PPj@wSp$ z_&jMFGi(+;(-HXu?KTQPyAg?r-VJ-ZX2jE!>jq1@!jwOcP5ILex2v_Z$yMy$(-?|v z3x+dQ*IlGnG}o4Lh3^8~O8irg&k^{onmgM-mJ|gk8Tgk{pBHO9QJYEcx-3BBOQL*b zEK&aQ7rakIuSn_5+4Sz!jwgiOxVZy-!N;|oTkRN`Vz@z)+gY9p6Df^wHM>VvmpHYb zCIJc*ab_jsrCs*H=()ZK#k(c2N?y&rG|7pO#SFrG`3Uhyid-(DPFXt0tiw%1*t-2D z>YQBwSyp`v*b^`iX^UOX7WPF2F$7Fs-S}}TOm=A{$xhKXO(*5^%w8&*2Y8P2-C2M? zSEXqJC78vTS9BSs0&?L$zLTHS3yy@yUp_KCf{81y3;dtYl7XFW6>NZ_jM0Gj2e~G_ zWaco91R_vujK|XBzF;aSSh+9(3$g6GHZ;t3!`LX}?lH>KGmE?OaC_1ow@;pziryoV zxKcz(IL5Z%oW;Wc@_o>P3`c-WE7RM!5IwnSEd05$$vsc(bPi=nxV}`wiHTe_{Kcn1iyjfim4$+YVH<)I*1cw*MeSi3 z!Ue~qBePW)W2dCdKZ(P?b$MqxROtB3&o9dq!I$y0C=Jm&QTM?8LhBb3Hm(mmpn5%G zpm+g$hM!vkDqD$ok_#t%=0Psp0BWCd;VVl^k2B*=x)}E165kh930|$fIqyYRmORec z6;utPZ9!u^-J`ZlT{XKQV@L>;jFWp&3tAjeA%O#yEmVP=+vd(`aNUf{B9lmbdn!KW zLOk3BgU6Nhm8F77=F)a`RDU>oC5R{>MvSefLagpM4$VbeydDXhzj&LoSdB_;c@GWQ$cEw{~~>LfQE z&KjSL4_CPyGXG>ROr!L@G-{gj#Z=s|PR3G5<$!F}VldD2D*$wlaDhAzd!05i(Y%y& zpG7)JPx^M80qMTZz^o_h@aQ1 zW-9aemJ=rDi2KnbO=Xhvc61q}z;k^d+jcs@QS06JKli@>xdW||G&6*aAM>5&ty@7H zq~E^(uU*?yEQ7lGlAi~S%;o1_97ORcKfkh8K+!us*?wDp!(`l=o0A^&49;+Jzd6pV zc8qK*UnOI3Ohv^791hk24LhUq?It9d8*Uc9GBSDj)cT@}nz!Ib_ z6i`9CELqE!s63&v^m<-y8J`$zk)B13C%yv%pAPu%|DIOJeX9zTp-vF%LYY>z(4E3E z2f0ACEaB00cz(Z9Rh0{MSb0JL$Lbdt3R$N*op5r=KS=s#!-U$l_D&-n5x(xUEiWcd zJLm=#@qU8-EfqBgTn%QR{^i;*<9#J7m(^Ls=$~{)@yYWNQ$8`=A_mBaUO+MHJcWpp zvsB7ADp`n~UUUhGi&u|=+e$8&&bQMED6+JaQM;mmi4Tkd=EKG>$R*Dw1eeq@xs{>P z^3-GoPeS?j$u2%#yj%7ul)ti3u-r~UL0AQ$VUoi)r5bWbZXV}CWKwO(2`7kk4gwMY z+wlwPtl>%xGM!i4N8FZvymX1=Nll+f z+Q`u2mrqw7iKnYYPHZp~u8e+YTMpJ<8&JF#v+sbu2J>qQX0{Zh*L?de9kKPMH9Pm%o9g@PN)@l{RI*ij#Mqk%!Z zz;u>qXuQE>nh0)SREv6miEuV6!gp+}Ed-hQaj9R6;X*1Ua~3AYb@Drg<|d8-4dx)}UFyc5 zd@YahuPtwQ8YuJ0^HR8agk)EXZ2d+x1(ZcS;k{6+jcRsv&XAZS%hRuPwCAo3+9qSg zN)x64)TT^No$WhzO?iT;ARpvH!r3d~ocMr$lx250kSgg|DP2O|Vm zV|oN*R~HMeh3yY--(=$*&1lF7PWIw$_m(5K3&I-SllbF~aI% z0T@d!oAflJv3NLR3!a#05G#`ozoX1T{;v@(y=;4yrp1xyi$RiuP}J9*SCtyGz0n>$ zGi3sV8|A_rdrQM1oT4XB1pDNdFD@-Ue6|-$1M!K_)gRr!q>7n}g<}8@9+6HOkBZ@@{BG-t>aYgl~kQImrH?mv-0yvfybT zh|Rq4mK&%JL}RrrZ#@XXTfpkNca*C9Y)6`5m!Vz@dq5(9FMuKe(*_z~NOLz5WYO9W z%3l;WAA?vv3}r4cg2a=lf_?Iwgr7(1%4!i8DpN20ImQ1##cHkF{{k+@b;36;_tZlO z+SBpXy&>ytLUs#~^2kBn0vs`4F5VIJ1%dTjEu!Cv3fOGbR zs-Eg6?D4>t9!-*s`sR>(EtINR1~Sc-fgU6ic(f77SVpJ4NPo%rv(DW17Z(3UOj{WE zH)d}OYW6w$Vr8%jyQ>aSCJu+c{ko3`5yf=bY_s6W%7F-3%vdciIkK^A zBd0`feLu5 z-9z^(k`TSN9=N3f2t2+ZDYvFPvtZDgakoaF?83Jw`XBjx|EJ0cfrDtBib?uGznYn zu_JdHSKjv+6Z7Me_l+kr2Et<=%orSi_u8Me>FQcR@$z%JQwN{P)D9_ScRH70)2K<* zBbhIZxd4z$rscy9?GTfM+je0a8Q&8oU7jsnHXU8)I^{bcEB+_eQEA@;NBTq z1w#ZQ&#o zu?~`U#^o-~UY+nu!p))~!F%=#**s&8486L+%n0y5XyLqImJ0Xm8qSy>f!zVoxV@lH zB%qj_;<5`8XSqX+7C`E4TwY494RNBg*fNw>L~@U~PY`y!;1S=1fB~>!Y(hP&4e`hC8w z#Rz9cmn9!QFWY)Vl@rjmh6!ji7q|!CKHCsO*vr&Mk|H!aZ2d!n2=6L?` z8_osrF(CDGnAHNJ&a;tYLRtGn!lx9Z-H8ntt~ymLXdO`ShfQ|WnmFPa1u0U?{5mT7 z<;n?+4ZN@jh@p9~OmH#e6M5YzkhnO(VwO~6!h>5#ktTZr2MQA(s5^ zs`@9}NY4;>^`=fwdx9KYHXb}VCJ+xxbFk>=;=v=by7szg0GJmEjy$jrq~TzVu?*5A zN@nZk?;)>39FyylT)OU+2S?~taLcfJqar~yqbv|Ileh!w%r<)!XB4nZ_DB~>BBGoa zh2RRD01|I3OF6k&$;Q@yqT6?(z$5@pEM3Su(GP+Fcai|pBgsW^k0;_9_1)3O>+@;Z!Phs;p>?PHHn??2? zDz=eRTAJ-TL{-`Zy1UA26mU4R7(hHvq2kkrVQsO1;c;6OcGu;3r&?}@~T9j&2ckFq14XJA{frKfpCb%QiPq7`42)Yh=g`qcAjA|W> zx)*+9mDY8C|lE*@ON#z06RG3&1INTvRd+GbFezLIe78Lc+n^FS41m1|8OP<F|lH2(l1M~MaJQea)tj)W(dzz4fuqyitMt^6Wa1-U#%1i{p5X< z{}`9yn%Gs(!hRAvQLFv!Gl=f3HtH|>8@64Dt>feAMos>wtV;d%t5HKxeufrOPy5IP z|MqLG)dN?d-ri2RZ2E0D$fy+cus-Dz8TD)iaWA;$eeyPE4-YCKpY|byTtd+B_RZk3 zbdl%Dq=4)&4<-fH3-~lCSSuotOz@GTD-*5)!$yOTT0t$oVn%slT2c71apP2-3E3q2 zM(Kb`BbXcz9bjCF_SJxta0J{Q<-!7VY>a!X;J674%VYb2H=QTEACh4|O>ThoJ~G7V zu>RQ=jKsDC%0_!ZBu;a=jN!}jVDNUl+_`@~s&=Y6+G!RH_be>aB9qfI1)6&yU<3K=LJ33{>b| zGNY7SL+v4eq`E`4k2v?)wjT5VLV_@%c{#LA&uwT3@^9mHEO^T%j^!E5Q=fz}&V_@4 z3lEPN<9ZPZ$w0%*!f#-g5gJEHt7}-)8&k6RV+C5exT*>vb+lTqc^DC>TVdO(^WTYK z9<^bM1XocQZ3eJQXm?FZ5o^D9bU8>0U$9iF{j8@8+>kUpnHtz9&k2*=r>VjE>w*#D z9%QE5v9$a|8nB|a9U!4saS$%GjKqMr< z_5+!sWIW8`N$@-k$&yESa=nOxb{d_&(b_d4l-*X<1Z6sa%CwOZ1ra51;-qx-OH$_d z6^LNvu>6(@)Nv%n=aKx7&g83@WNt9P!;C=*^F|;m)s(z1jCE|TC;b_DD~?a^EqjKZ z90Yi#a|eN!FA+VWnCnGymZx>@{W}OgJiDRE?`8^?BdQP5RY~Op2c+C-{WOV(8Q`hk z&H7v(^}}GuNOJ1w-nNAkC3+TV#n!RP;JM<$muUb(vh!dX;3@!*Fy?v@NARBCNLCt7 zY`Q6pXf*{Z&P&@@{m+Mf?zz)(9D+^03kCEW&kc^Jc6g|O9dKOy?9qL5bq_hcmRe%G_`~tm(POm#(sw!l%>y8?+)T7KO*jm|xP$`It z*D6~VT@aXCTvev${S3OjoUkV^0Q=;5i5-uq==xHD1n1FCZF@Yq?~`&uTB2r@R0$cm zVoi?mcRFSc%&X@?%)(8pdZiuzVh$Dl(EZ3nAZ#QHgQJU>)kh3Fk^SY4)Y=V6C}od_ zHcAv7($0|V##=0$^c$5k{vR9A{5G>60Ip$m1IfPpV;b30g5(aXVO)fSJs^tt>EuY zyH96M!sjr}Vo`g^L8lN%Zso6qh{RN%eLz%P7`QHF65IBZB(&-rVXi82D4@Y=NI7Lg zV`+w6@aF;4d`kSUuNGVo@9L=2LGF5&fPJ`0n2J-=bU|^Bhc)Ns*HLz%5AI+;2GSX8 zo;GNAKrbcT;qXob>-ZWjz5$~XfQeNhA3_N-w<^#FB;cEW)YRG14_D9iHdoc|(@JwQ z>}G{5Hm~m@PT%2kdP}RypOy2zSiJRN^@3~XpmZDU%SV8OFb_TgE+P2DZ;J${P8wm* zeUnXKv+-Ldyn5ARj7r=!RC!E4@l>2vlx7{xev%4CcO9{2ZYDjon^xpBL-k=appL#S zEe7GsLV!@12MYlg3O>=;^#Z1GRxpfC9O@wFWy_aIb!6_Ftz0_F@op-=p-G5eE0>Rr z;&eL@2Y@7JX~Hs*3knj6r3_w8E>RXZeW$R7LbebHiCz*cxthoY;Aq`=S7sZ z{`nU1h2E>%_U)hk21K-utJniAhOA`}Efi4WLll0&YUF|U*irCR3J<~=K+iNaR;D;h zn@Nn$5W35ufccUq?@6eilV8dF$6j(^YpKHH=Gewk!O4%hM-62imwdEMP^UGNJ~p24 zF#Y7pVP$~splsh@gttFhAk;w(S}E3pA>HCdO05Xi0KHfc#211E1Oj+H4BH?K`Vs}0 z))!C++hf{H+Cyy}^G(^%@ub;L=&a6xw3LO8$3A1D2=_DwaX%%AfOFE5o}`0jLTu!j z#AK#aZ%*GT!XaHP0vnxuq@z%5V1x`&tVI4aW{`q~vkA@htEV#ZC4?}*f*j_Akd#e@t|g*xW+Ri(~nmeW?}Ts`UZ+b6qg=EakZ$9806xc~?Ok-Q5U#vhVB zu3{tGLaiUxj;7<}D7uNzvaPqV^Bxi)e1QNJDWroKopiGCreYYQubbsWTGpg_?*+oP zPo9$#_elHLC|WMZiJ0AR6aJ_*Xdy%lHZrJdHXTeQ^l^F}!3HScSCbZq5Wq8s?x~FO zfsu<0@dyse&y*HsKxuj!u^gF@S9)wca04Aj4IHSx{~L0d-@b)7tf~R(Mo`UJGjH$) zbl4kU;F4N5F6jr945*Fy2o!=Kla{Mse97?>V)G!!Z`Y7V*2uED z6>Y4WG)NpkqPeWL3q2SBDEPB#OT5Tiv2c@CP0&cW;E;Hb+sC(WL`d0ym_=~^d4LR$ z)QpXyEk$Qw9S>bu37##?d%V_$!@V32I2fm1JFy!=omd;dKSf65W73P8(-#AHb>3y3 z4k7MjmE^me_jRDcQENo+5}xWD)Jvpk}NWc;N}1q8}%Q%Vc4wFh-h`vT`0f(dQT^A>V~)=gHvk+-;z7* z=5B_Bg8Wx#TTz+23;JyUaaIcMTIj^F(WbyG`*eHKzb7hk{reZMA${uKZ)_A4kya^8 zG0uFOAVB4!#2@&$Opov+FqhXK;B;nDs(+p`ZNVP`$X|cKL%N&B6lshI>7TSFt}@LSMm(i% zSML~BzUi*!26>ruqkups2ZlTWozkNq;AStSbvmTMVxf;iVKz+)CQECU6l>@ED8L}C zEIy3NyQccNs9r)#fpxA?fp~NK}6g8{{MEv|NO6XRzsYR7#|}BI+>+8S}g&Dz;u#(08D?=s;d6 znIdLVW`EM9&nM3bUguE~o^vm_^xt+408JnOCS0it1lGwBa$*l<=nlf6 z0A7SX1ieV|eZaq7KBki_(vj8!Itf<<-Wh{nRD4(+=vXI=izyR@80lWI!lPyu ziA)m#Y=M=E_c)JIC4SxUfTOrH7K2QW|6yp+HJOxktJVJquZLGMW0ydMnQ$M zt$d^MmfjVN6kVz0t6fPyk}DCOCFmLy!d1E?7-=3hc|8Lyu;wG-sNkX9Q3AS1s~;7f z=Ll<->vU!k)PGVu@X7O1@%jWui)N9hdsS1OIE}u-a3&rc>4D^tkxsJJ8jF#vbsjC_ zOkF%0F;eyt16reDD0z}1eLL*b}TU-4g)2*0QC%3R3%yBozL z-nvjEoVpiM9GEBO#JpPo2-|beC54doZ>10+MI*}C2Y$S@M2v-&+tNWOq%};bR1GWx zSDDBY5dk?$E(b!hZO{_MqYx@&W$=Ia?>Tbq|E!6P2JP^5`y;vICM z=yLRLG0rk|4mS6hc=Q1=GAJx7AxGWGd#SanRbU|caEF48I$k{#KxU6Qb`7p*Xx;Ff zt8o>2rf1$56N(bD0n<18*K z>i)|M3kBHv2qHT|C4$=^Gy4Wk+X|2#vRIicIwfHx?2I?m2{)ZH?QTZFpQ*PxkABX$J^*I43m;8|*34j8)={)G) zR5|UDPvS$5CT0Fp91jIM^L&jB-o#vErpnT#dsH8meMiugVLLjIDg08|#?vbhoRDC| z=N4J;_yXSICaA@ie3?|(po7TTVWqp|qdSXC{xdzg9)ZGzg@Ueaj3qPJ7Bh|!CgY|{ zm`F}pY0`OmM2WDO#aY;w_+(W;QLi6Dow!X0Kz;6ee!aTGwcp8kc24rPvdf~8GV^pk zPFb@LDxY<=v&CR`Oh_|`IL9I!@BdkFkrM~xi*tPh={X;mmpx@qviBs0)p^;t_9%P5 z%vRRppul!*yY_~NxM$8Inw^Nx8H+Bv)*o#egdnzb{HpRsZk#f`*H5HRnsX@78U%pY zPtf>k2GDlcfkPiktpnX9DS?P`zx0)aV`XH?#PC;9#&^Y+N4RgH$n)H$1Y1&jJjG@; zf)T?hWOaLub?2@|+kIou9tOjNCto8c_{>%5E09K}pYU3b->B5sNs0Q_0G@N=^n}*y zAp)k7Uh9a-Ew=$*Cqq(6*d@0xcj~iBU(?Vb;2S~#(9zM|kE8JWzm_Mwqh#-?h97C< zzxWhk$s;*(p-64x0I@~1wkpn;eS{4R+W28WjCIQ_qe9KDqufxcw&*XBbe-(?t|aA1*<^Fi3_GCY)ib=KfbHgosv zfbZ=!O;ry2Tm;uzs^*_oE}*4xQ`u*Vz>-_aCQc-PxNHzCz4nwJMmqQ}KB!*uNVQxj z+GwsBL!MO(1g1k+E0ZCR!NLL{G53zExPO8TyBj9ZI>jKrkjCA_RB(tw{Y$Hntze>zX+TClu?4(eo}`PZR}lTV`g~;qEv8Ie2yKKPI;HNpvhFAs=D3(}uW zm`AlDh_N*Z$k!XDqp^St#BD{o7KdN4Nh-^-oJp$wc~07+V*K(#k*MoPJrT0vtqHr1 zhQ99Myn^s*RL9B3*SL@qlODabXdt}RPP@3Ei6D@e>siVpueBJ8P7t`VPogI|c&_mE zjPdy5)!nj3#^XYfk9#J&Y-`3C1a~KC6Ah;+TL5fMk%KMmZK}gRBnh-#w!?1H3X|KI(PuCnz@N1q}44{UiFWDbWazDEvh?caDQb@Htg_UO7|vPjy=YDA#r7EPnmVB}Y| zZ6y;L`#q-@yq#z=(Jh^6mVIq(kM!SPz5wxwQmz+}RP%k?DRItPqo;1dIU8y6G&YK7 z5$8t))*4*q5PdZmXaE#3mku^aDFdam1is2<8h`&kxQCEh>wZhv%ISUM9Xqg~Au=Jm z#y3vK5&P9apZI0yRaTF~muNgU)jWvCyPL`W`A`3&Eox8g z=qQ9xkH*{V>u5XYcWn2dzX;jFW+Em^xS5WkHZas?(9}1F0l97TBMY_&Mm&@^QeDs` zQHd`D8HUVQ0R!c$QdimfY%=msxXSF4U!Ls*Pk+Iqe*N-75#^N?&#wHL%Q(`Jwtk_Xoq#6e6G%~8uH6i?dpgviny z=tEgT@z^*mE)h>>DcZtaO3 z2_*~&TR**_j1HoKq#_j{Ms8#M5ADIqN!$;=^lvC%(43^rE*m1AMC=KXc@VL;5b?N{ zwYX3q3CLiR?)Ns5bBak>T}idR&On~Y_#o(`6QpWoJP{G>I>pQu6)#`kG1-&pC>l$T z@AL+-d7)@ncg;R%)Q`h!=Bvr2sm;Z)+M@^MFe~dc=S}&twUd}NRp%%SU~USP%bTgJih)gs+Cc~$?@Td}ssNn@zl|DF%3^lIx6Iye z;eIOETzfU*dhhGC=~)1tJTyi6(IZBW)6rN%xSD@o2Q>sDZI28`NOlw{^OkjMIFy1xfJm0 zs0NRqlUmbQoAujNvSjZ&0qET?SQ60d26w26BTFG@MKHZnF}nr21>*e{gtu2RGhi_y{iaWDJTMlb7^~Q=G;ScQfB$z{ z@1uwe$A9Gkm&w&tCmgCQGL)EhPV0(&a;f8*qWswW&_P!DSYFi^w?B`b=&PVS8qo$U%>btu82- z>{y14$jQxOr0^stk58VHtIQ*rX>p~Xt=bN?;YyFv8W+3G5Q!7*C)>@2^cKW&0KEi7 z45CNs7*T+`bft!pKJ@3QC_y&e)|i-8f29AufS7897=P-kvT@+ah=B-OnFC=r%MBil z7?u~;3N{w%wJq)=kr%dx><#pJ(7|8(lJepoMsm_pA*)Vz5!&prHAN~?)OCM26X0u` z)QTG$WM}rK^4ic84C^fJ?-%8k-%vV;GO=LmH zcBET_QzEiSxEUBjejWYPJRz_+CFh7XITbvgBdHxw_R=^;!52DDtvzdR$}EgaUOuyU zI6Uq0S)T*SluJ53M*~G;$b&8c0U=o`!96x3Y!ANwM&rgfal&2BX>d}DRtYB?Wf##Z z_t?vUR4pw@fgjqDq-51l`&v+eCsD3-xjcBG%iAZ8ZoyT|Lnr1sg#FTgJ_e)+uGuco)DbMi9^AbKDhX+fq3yPasH=qU?X67SxoJ2!- za#_6GPdB-6V68vwX#6y^0;kq!5Mqe-orXvkHxPj&9yNvRRCas{+hWG72VdmHAUD6YB zd=iHYvfGG6b#K5@YKRVUL;?GQH+YmiGXb2dWW~?Ur+xCgWEqdJ@Y3soDK;XtUm*Gt z8De-^I|Si#EDn+D(sH^)jyc~od0e|q{T6ktoDegoy9YOvma~*8GT z(E3Fei(NUt{rcOlhoZ;WRC@H1T5jD%4zA%UA`|^l@6|m+cZFW5zJp!C2bJg2;I-ka zR-OkJ4|j8Ugn5^W$TXX^bJW<`B@s~~XOf>6Gs&|VC;VP@OMtt$-@2Yj^GU=Dp^5?s zo`0jB?y<6zN&&fC>hNXDu|BrtczLYFBb>WbWb{^p@{(xBEbj!qb5^0Wu>5Z!*x_THw!XWJnC+rRv$U;j=}94ktKbyKxcMgdUsleJ62 zTh2wxD;XY(ew5`1k9lzKu+-qwy<=&y0E)AyKxQQTuNXewvYD6>?Bd*X{gUpOju|h8 zgkn`{#}Yw7(QLp#pfA`L0g}fNIA9#0;;TwcJJHwIObX~e!nluO{ww6qJ4z-8Uv$B| z&CHysjbUwsjrIU86$9*K$dIQ7QvNSwjW*>x5Zuarw8OEpC2M@~QTaJ@t1?d}6 zZc>)qqb)d6ddu5)=1cyx3q9Cv+Rdmt|&FTCs&v>!`p%5a~Y&S{?L$y+H_RE ztmYT~Hy=o1Bfn|Sp07&g7*D1h1jtH;>Ark`c*Jy<77M5i;1svcM$B0KPwx)DdNY9z}a5j1@IhtV}G=Pu)ejVBmRErhxYoP>LOhWFo#$ks2XQ*uMdG zydVOb*uIY12DM7DL^(?>>=d-h2GJ}&9?t`UEP8}pmx_o^bgx~>RPdK5nY`^}063pP zUc5F5ionqZ5W6%UQ@`@wR7kA}{yHGxHy}&O5l9LcQ%4Ge66H}lNho~g`)KpydfJRo zNLIhAW)>`$8Q-vqK-QNg^5<;LTW&xKTZ;g&S$Yv-=_oCW$I|YyxdvN`cTv{xDpy^l z{w3%TF9+crQB>#G$uA|7h$j;ck_bftPFuIGMbcNXtxc$=hr@wFR1oVg@}K^O{{_+A zpZ?|___NZ8gMR-X|3I$2;n9Dl!1D>%7FkphO^uM@AaI9MVf&-O_@}gLlrsXi^XeHX()J(!hwo=843c zXbq<*Mh3xUWxE|v(*WrOF#Y1nd$!gP=`R^pzPvovUyr;zUYvEe=o8dkDVPILywuO$ z+#_5*fKIZhfa5htj9)|R325bUU1gCMLh+Zmio{(GtEZ5+_Yq?&mtmGY*>(^L^I+TI zS|Oic?aI>Qb?jCGfu1}H3%eW2XO8kxGZt67+!I_fn_S0*sxzymqS1TZ3#Z&(bi(Hb z#Oa>Yq*2t)`XFr%E0o5ksQ12jFya93?R$~16y^bApJQU zcEeljDG1=Hz5Y!E%mE9udj~zeY;>&p!hP{7nU|W$C+@pi0PfUHhrR+uDrnriVbhI@ zq&KXFT9h-t3NY9hnJ1-Nll}kz5l4-$Wd^^NStz-m6){OW|V2u zN+#i2aA;63X{UkRgQ$^EhO+1{UrrnoxerIOwfpSGl#L%x7A9nud9W~XH=E=~n@0$Rw!LiV74-q*m`1{36q*B)Evf(1IixHh7vM5a8hv$%g&-8!~?IP*}PoG=`PG*3)_7+R7}3D^E*J%bB#ui#JV6J~8Yf z*}a`n{cDmBRiMRADqip=W$z_QXKvrXW)Zw9q{n*vdy|T(Fq3HBw|0aWC;Qq&0m_`p zG-M|2y*JFuiJn~BPmHjjdTO)Y$XbjbaNA0g2AR?A%gG+tq-;vYk0&b^vQ%#6^5XL> zpH?myr#}u9orTSm4jE);Vsu5{^qwk+9DpjsP@dvPsTsUZE-DdCdCQJm%%A8J`gFdO-#mQ6@5{c@`;AYn98a(}u{rLncxPNRH(dO&flcOjtv~ z!YaE}*jV{03J$ms0rZS%1?v}2%-@lz%*&T5J`v!Jg0ymUy0ryd6Ab83NDs(#isq@x zgFNCqDudWcm zB6n|qwe$VYcj32^m(J=e_AyT?KKSH0nGY;^#CMlp7ZkF_^{p}f9(P2xNuU^*U|ea# z;0TpFZjZ6?uw)q}Qa+VHE}~y(a3m>4A7u4vX^G9hj;D-W{XRWqmR}dpV3v#yPp&dVNA4=~;&-}FR~byG z9)}-J8PcDu_2}oArt>XbQ|IAvthzZQe0stEa2@(Dwa$tzN~20=fjG|B^8gT=kj8K z*4JBVejfVR$dVksO zu8*ijL^)CRI18UIe_dkOsmR%T#IeK?q2UxM#H^ydiK>dHq>eF`EO(I*pnmV>U`a;)V+C$CuHi8b> z(l*_q&dao{${v&_e-s|$dGJTE2=Q3WuM|N9^X|cR`D_CcdW+5RI4B z2niWA!?yM7-yF$D~ORGI&H?Thsrecq+HG9x0NWB4o}oh4_rH+eG3cqDTEwX9O( zVhW79HWebLV2waTX^&-882jhJpvwL!2d=W=Ek}PA(z%pO8J^4^IP`fif3QRM$Ou?j zDnJM7d({uM9RwAuNTher!rr~@_z7JfQ53Z+Z4|{W;1ZWXLYL9kFF(MLrov)!VdHbq z;3|5q(D393L1@f_AB2U5Pd|v2<$^`9NpN}0QD&fU!iKyw<&3#tC5ROG{^cS{{fy6b@5MOqjC4oi>R#s z+PHh$!hv#V=e3~h&EEmv6Hsg>Q%n$&3SSKN&Y^f^1yn`EhDo?9Z z>j3j)ppT$CQ&5tKj^vcOEj&`KiKAj1=_7H4>$4=BR5pjt;-T>Fxnl#!~YH{TAG?3YJFJtp&#!9&S z4Jsny?h`@c;l;7(Z<)4{L;U)cOTb{6liJ)*_JeP`s53mch;e-OsqFG88Y<75c-}kq0z?ZFAE+y{f^F^r zkvv1`Nav#Cj1I23(jA+Ovj3XO|2gIeGnorKfdfncEoIDs7YCmfd;*W_1sL+?+q2`^ zxzwSZsw;pIr5soR_B*WU=iOfxOMSN*IVtl8i43(E0zg+-W^&p^5kd}>lcr7@CF*Tc zA@(`ab!A90(S&Kc)gUseFb9llRE1qIv3!u1Dw{})SdlW}@gyC{(#B3igC{El`{X$R zITw7Qoa;s1L(FIjdnvz2NMRX{&Mq(x4w#-X)0) z1Iq@FC(i?>5d5vX#s&nHY#bbF4~5NEs&VRnO3m zc#uT6GsJ7yXUeT^$r|61=RcEg!80|n9s$ACBI}5quMPIXaQGY`v4+9)uD>p~A9apN-(W023$& z!oZo3VlUF_Ay=}s2E08$-83uX+|PJl_6hs&M!MDWZ(j9sLfqMb-n~BJS zT_ljv@|Ti7C<5Bsd(cs&&n&{q?QOzE&YO4K8(}#l%_RQ*tI{*uOHF9#dnv|NER>#c zV{=<~!nkbscvAV_7_0oh_=Uab5$;@FEU5P0a_y4%%?w=Bu?FRvmG=|)KOE9t^i{jZ zceyW>ffG(DrE1E{MqbZ06z~Iq3^u+_#<1DsCM0)}gjn3$rY;yXQv{GY`$Q6&zR6jR zfEx)kC-K+1aG^*~PL@#F?^$YtS-b$AmZ{d}1wHj?LaD%oCL5deX)~Gi*=^@G4?>3ei*K z2^-XOlBSk2Wuc`g9GkA~IwK`Ah}=@F76Tz;8pFxolQ`6uL*&)&m9UaBK=Qt ze|2w-t7f`-R+14Dnj9I&a-~+X-LuBYH9Fo3ztUFR6FR zqNwQfd+O;&4gfFTl=(zaR|>|%?@UrA;}VE`avqW$RK!ezc?yzyCUJ51tA4kYD!E`=x)*>U130h?pz{BymZrgTDgC3=W zi2OPyi2soa_H4_Wr!t^W~?~~@|L4V&$0iXK&t7`>fKp!G(^mt(1HvgTLxU=5bpj9^h)(1wM;;C7h56<$ZgsSM331lBf(2Mf8A;X6Y|=zN>p%$BxReje)oA zh_c?-OhI=o$=t9F!g*KO0wopIT>AN)InIF-1R&E~L|24N2C;F;;4)8ypdF!!8&`td zr+9+~0lXXJ4ZS!jjnXl*xc=k&;YnEANRt1>tIQ>zu@1?MU53P!2TF1e0D%&33Nk_$WRc5NLN8dCWG-~!Jx>Ibf7~&6fBU<$UrQU zvJI@bsWsXu&F?{gH`!&`v7aw-eR51LuK)7!;&Cy3wUCg=k#Wa}(&3T^0EkoxCbiO? zXkB!Lx`Cw|CuK-WORo8h0bjK)@>o%21EkhsOzaq7pHiSWB@hM6Zc2y{Eizm@&RjqZ z^C(%{pm+b{OCEvFwW9e86FgMP;C{+*bxiIur~)aI8%2v75j%Hhlf$Qy za1)u)rbEUKWAx{KWvVcteh_^yCGLmhFXd5 zw5tMR6K8|~G6Dr^yK*XM>$Y;9^va^+zLf6?grzy~K6O`>N91#@$k^NZ7H$D2dNbL& za)zdLTHTGp7_B=INvTe`r#R<0v>_aXr(AdA{3m&ECvg?2#Wg`C zR%NW{QNrTNDp}X~62d1~=0OPG1&c@Yb#0|!+{TbJhH5OS9ab{=I$xbqMPrd92n1@= zsML9I!Jv}sT33|RzUuf1GFhwO^hG;-ma0zCSNnmcuPv)x2o3pU3L)2@$h!!#5Eseojj5tB#ZGiDkQTNC4Oxxnto10Et}{TQ14-NfwD@4 z@=_ceEwP>Rz3tO}#dOrV6V;mz58=JCDdqP{Mv^Z@{f+rh)OV5O5$aqkGFQ^fwtjxk zK4;`0T4yYa-h&NOOiRZ!bbse2!c57v95Mq%BufeGLa>Ggxi^4j?nct|7M7{wEnDvj z%v$z7nn{rVdGJl2y7=p_3+7%&u%|@N+lB%v0DR>r*dS-G7Aqf(do-7~W$< zIOdGKAKdzJ_JdSwG-fw!D*xs*=wU{!@8HxhBF(-tk!&N;9EG8biM1P3_Rc^A-xEqT z7xMbC@G7kw(@uNOgwr0piUr$M*Rttfwh!kxu!$z6=BR&eZq+BCZ`D0cW!P0 z+Y+Xo)HKVXMLMooH1hw;XBO_09gXHCyG)A{j}657a>1foKHtvP+qQy!BS3L?q>!R; z7||><>dZZr-l;}G;LkT!&a%CaXdo=eY`zaQip^hdJ1j!|#2x|EZxbLNOVMU@l78g* z-6y|1)yiVTV{5R!S`Z`oE-j5C$2q8?<`&*crw>}bBl4Yr{2{PI4?0c9W_iRA46+T< zaUd1)%g78|m11&T#aaucr^l~}PpR4kb8a)RQLq-*gSrUviplJBs4+5@mIBcF)OtH( zQ*6`B%NHWzCw7E-Vbk}ChGoNS$O>0r|XM_(Sz-F&sxQ7}j|qtS)#C#|Yh z2YNp<_aKU$(GTc4lHnogp#O+&JW?FO9&|hD5(37ep|yL8Qbsy&nFa*tkM%=Y-bEsIMtRN$pVvHkR5o7kUrF zyh&1)k-L)HVCAD(u_+33^D!uz3%JGkViLK_`P%d=)T<-aC@SNG>mvnxB|$Z%;1OM> z8qqyaB&zlLG4H9xM4RSo4C&69=#i|>S%cNoZPS`*e!SCV*V26gUtcGImy2(bmkTn-PA&9sMXx`@gYRfM)K+$t9s6 zbj*nepM_F*Q+vuR+-BG{&R5JSH{vkE;Div; z#&HF&092*uT2O8LBK^j4UCB(Y08i>l_~dzs7LOE+jUqQIBs$~jN0&C@6!sK2bw6Wb zj?#=^3<#BtHbG$&-|g^{H=Dm%Ep8!2K4^CykFR1BX|sqyf0iKkGol6{_`I`w3uz|}MtHVRgq`m|BlST5l5Y=NhS zJrTY4F`Mb2X^$YvXm&QLVx+7?@HD9#AIQi~-9`AFm}EU+vLh9+{az|bD6K_z=#BW4 zJN|9u?j9@)eq2~SDXomm5MDm5cmzi`Rtf-=+%DPNh_;7aj;tXbA2FiTFiGk?+uN+N zIv>Ow>fnb_agn6LL|AYqDX^#}UmW{diN=+(WSxDjAJV6kTV~c%z2t?*;XZj@iW85R z=|&NUxPEe2gRfPm@u8YM>B+nD5TyeXR^qZD7IKQHBmA7H`9*w9Aox9m?L6rYE5}5` z2NcQC-HDPuka^4w0kR1fDK3fhrZ$uB4M`wG#Zh#F$Ln?D+fZ=blO3&eh<-c@+$YaV zcz6V0H&zSQ8+Rj?&AIRs+|2S`WvACmdVQ(2Ie3-=1tSaO4q_dA5fzLogqU!YPLd4z zwv_`ihH48pY6F^+$QCtf9NgRlh~(}DZA`VbTXxAb#=ZLxcUJ6rDtiglZONdS#pU36 z(2NC-sOZLOfejj?(ncGdqEl);UHoq;7$riRua{D3TGM#%DL#x4SUd_Mp4n5ZC%EpT zUO%(x+74RsK4gQVE1z-(Wo1vmk5zyk+uSPPC3pT@73dSB+$iWKEO9q2hiBk6BlJ-+ z&j{Pu0?&xI{d+KYJ?XJU(k66xzy&gP=}XSj;Q-(RFTIj+Fq37#lgd9nd0x`DKSXoo z_3MK4U28Out&t8iNnHL9G>VR@$twD(jEi~Q;=m;XCg@F-e-D6$PZV+*b6(I4YEDL)T~7#sW^rAwmIK;?q2-gj@^H?9x2bY@qmB$d$!<;KKZ^ zw%SHn5G}Hmm4saw33h)=guO~-MjMZqHsl!|R~_Lci@r&xiq>&*d5sYvgogq@!sDaH-4Kdr+@5adW?F$) z?OFskSt{2PcoN@NY|MlBzRULhU_fG{07^2_{?|dZW@AY?34$k>nxPFSl_|2dji3dVjdz%IhLF zq+~&=W~gzLbd!eYWr!Dpq>j(VZPuBv5LEO#*ie1Q=X?x)kw_(xrPZ4zHmkVWGI4^Qa&K;Qly(EI zBWO>xWF5!o=VwNdDw%FP>GX3E z&4o_Ciwl1oO8!R{#D#5d>~tF{JCE^uu&bpLHY3j8t&7SLiarHyejCfDSs+I=OeGpX zy>emt61J=XeZ{o2vujDUs%F+@Dx|Og(%!_=42X3%J{QUGA>%Ta`d$MwaMm zx~8PwYPW;I4a#St!=dpmLX=oh>>oj3cqi$DcduYzVkw4-JGtijpPTwm*bX}h7)y|T zJ*6fobwjZ>XDODywEfzo6Dhw*M&j zTB$>u(GcSv#R6R`T{mi+GzV4VSIg+{8eU>9 z&@GmewZvF9NBvbK{oNJfj|#c6P~<{cuLD}}SInqFVwiF+4fEC|gSF7zj;aj}HXhEa zCi28qoTRAaP93XE7=%`&zx|s3AG9~RCy5p0ul)9_iDhcqG4T>*h(;Z{Y;IkJ8cqCk zJAgx}peTF5emnq_9C8nUmro8Jk<*2x0+z+e+wC_MOg&)E!m`rlX>?^yr)qb=R}jQR zyuVf91UJFUup_iV*!FnCV=9zvP0W&m`WgT+Boc`j14_lZ&5ILldq~PgjKF_NqLzxc$|qpJVP!g2wtVmJKaJY4FWW*k;O$hB+_)>c@p_2y!hlf z8Qv^;#C#WuwB0jM1-2|bF-4~-aakASGR_lnjdKpsI=ga@l`GGeXFQ4gD-h;E#c)QS~U_zp_wd zV_FkfB`{!wTL?7=?^fmaP{HN^Vg{{5+K0Sx2MxC#_BcGQX1YHub@F}nm<0+X^2*-Q ze#H3o80KR9FJ1(C6ysl6D4fim9}2*Pc5S{IG2jz`!Dx173L+}@A5^LJ+<6~#W zk_+5H-$skm_Ac&03OtP_YP3x=a517@+&9)hamwvqnVXZ)Z=nxrD)?{c6E>x@_bmGS z&oftF_P3(V=@lrysKMW>9a~kEk=k6QbY_hOL(WXmUS?X<+hobZQU8{CChfnfzcS+_ zEE=N*s`ue#S>mfFfqR}{K6y^+>}8KQ;Nt578r@#)d@E-ESz5TJI%Pm0vj+gbT)C$@ z-n5yVjOVwi{kER+sd@`xDb?A>YUo9Y&zk(bFeAjiGwD+Y4U}v0XA-x6()5x~o|owG z2r(`et=oLqGKOUhy3HlWT&dk6Tph4I$FAfg^7B^P60+ub5Cj*w=&d$rs?e?^f-lD` zk`I}L2?8J?19AQ=dC7ZH##f41o);eFvPU!!or%Yp(W7dPL(dF8<^_-k2~YU&rL;L*hPWn4>{Iw%r0iL_(CBn94}C?8N5qIi?UeGlaY`nPS!D5_ z>1OvxZ(Lj`AhYMPl7rnXW(50*+inSP*Q&Ohxs8Ou48v7_ROYZa!$|};@^WGek~;q} zL8bJGtu0sAJPOmVEEcKd?Ddj5^Em#I-d5}Mv+90i8K86O+~y)&IxKq+w{s2i8HB_e z)(2-^G*dV5e%NJ&VLyRY)U3V>yXY`Ux&ZM{x1s(aYQSWIlY}82zzEoC0JL9Hsgbkt zq)Km}{PNea1&POceR1PQPxT}6ux{X&sV8L+e36J!9aUe>JJaq{7r;ypcA}us>}O_6 z1a*WdElGK+B)^P<^CUuVpX@^7C4n#t37_SA5o^)ywvyIQ@WV5kvpc4B92n*n$R0RD zgKmfwZka2WbMmAXZ=XCTNx@@@zVx~vDSQlr z8_vI<-oh`mwry0e!5Z36rAw0OBt4k5A&5cFXKwZx@e66hHxnv-M59qCq2SsKoU9GQ z{bo5qEDp(&M7(|SoFo8`#raZEasDnz(exhzx9E~*Ad2ZAk94mQ-VQ`*f57XRDiO|- zA%jt^CjYVWpZ_Igmin!-oi>r&Wk_@*(WP z5sBV##Db#Pk(4nXr0we((Dw$WZYb^%^aoV%!!#|TDQPt+BX-E|YiPkM<-%$9zW;BnR7-(- zy-*2$ky~2isd5GEOj_|zx_S8IIms&?NBK)d+T;6Ct6kk?Mc*dH#o#=d`SG=(zsjkp`GmQiT5q#kV*NnOjCM(PlV=%7SI`SjMZQQ=9Fo&>UzsW!dj zXxRZ`umA9;?UT2982^c9D7yR8-{gTZ2`S1Vbv{H4e3{_~Rkid}oQj^OC zL`8ync&d^_j2>xBLlEktxM5+W;R62&k^~OFJjf&L5_4w_DRi%{FgIa?H~#)-(`?ce zgK32d;vr}m!f;84gmM~OH_I0$YcoKYPfH;4phxdQ#p5p3QW1EX4JEWKMYSM4SOOhk zXoW~E_t`syg23ixKBJ&TQs0A>R>p5GFnm4;*{C88#AwIULdkmX-^=C{Pnz`Plw6bk<#UQh0Cj1xpvf5_BU|*ZU6YW$f2TMUwQg?ij!bBxuo8nU z=0uz&FxdCQ-nWP#s7X;AV?FMeG`bedDn{8J*@=Jwu&1j&?}c}i$lHO$NPu1JR|Mp_ zkjztS^G~C3#(((TM4J6@VR30~L2B_oo|8N&$;^cgy-PA4k=CW6ZJSAda5oL_hOE_) zL*90)mUmc~sNNHw$kc@4pMSlLnU9;y9BjC$s+PSfIen?1#W-48^y5ZHeHF?zH=r|e zMQTS}M*EVoDH$c6Ws8i)Yjm8ZMVl!G@x-#ofCiZg6qB-Z<+i;L?<%w3i#A7eGbKjPY0Z&r%#K6kxob;vRvyWgJdwg1+g*=PMz=AY!|C zGaCAleHO=kk2tMZ$eEVxFz8F^LFA08XX8-ubkCx7KPqAW&>jqK`)_{wiAMhBRsU9Q zEn)8!`H{3rjs+RT}peh9yBEubQb)AfkEYckM&1DuX0s^FZFunC-b0Q z??TEa)VW%qF(th^iHhc7h07wwGm<7HvhK-ESzgF4kc@Nv;4;{1Gk7)MnX~_&ySMpm zW8KpI#<%|p<{-f71`d6;E%_7OC!wq_TOZbzwC!U97et9HlVwpXZIUIK4la;E1{oyN z02j#MT;vYYgB~QAb|!mw|0DTJ@_W`>_3}e3TCP&rgn;g|Evi_g>V4O{)_T^no)t3v zYCsY8b#NY^(Yu6g@q?JU^v*EU^Gn&$7A)PPY80aE&MOioR4Grt_B5vp!x1>V<_|dc zVmhYzf(-=b+tMXgN|}j}&7CUy(Kv&rpmd3xDP2n8RW1G>58@(*6xT|C*He#C_Sm?M zME%`3gesNd*Pi8ga9U+{<4X|j7c{L&Sc9KxS^vz8ZjW_F)JJjMwMa}wbo1`|9h_XH z1^Y~qki`_y(_O7xMY8hP_A1EA+ks!D zto+)u9P0e$o1?84v;C-VkRYXAAp?d!;Q+^L692G%7Bg-gDJR&zh*op5Usw_#l?}xp z*Acb#Ia7BHR|vph=>0E2c==H(O~=$Mc+_RUQRA$cWO=mE6xwE-BJEIB&3@az95)eT;D@5ai#zJbb zfaEAI`|!Qw@>C@V|8OD*|HHY`RSLrM{1!JdaY%N$8gkzX$kzyl+UVS|oWkUms0XBD z#XM#+w>MrVL$M3Uu>ol}pd55Q#SAbPq)XT1h5fVaj(GIs>qVVAI!-G@%GMv|9H+Yn zKhvcIe-#dESlgDbYWK#J`#%#6lRmec6sAyL6wc_)k|N($lE)W7h!nDEUMH|d%U9)YZEU-8Smr1xAEHr(^BeCgp85NT&`Yja4*4|$J$15g3NJ4gEFp_t z)lSI-*MB$}<57iR*!p^2ZbW;6v|%=*fEf8g0BVFbEib)b%$Cu~k&>Ae6d`^S((cyVIpY;^6wT9I?~JR{hOs-3}O!BJdLkkHprgz%dsxlVzs} zsHnJ2ZDtUiIYo@Q!$IV9r)1Gk=QKaCgVOi@ev*c zrL9gfrZTjvN>Vh4#2D4?*uuLU2i=kKB9)g!AU!{idcXhIvKyr${dK~TRnT8|CrO3u zK|=P$KqWq=MXb1qh$Q6;P(nHgE=_lTZpQ#{ zU)0_`5Y*NSiAzsl5cH!q8mD{lHHpzV&L%iVKbwCMrLeDpq$i|w-0tC?Xq2JXyU+BH zLC`smf_89TPKb$xetrvNv)Ea)3i9l33sniJ<_)^ri0O{i7Rtq5Cc_hL^MKcN2ysnE zc>;#B6Bp`piiily0JKHaC30whL46nziu*B@6(M$U5lzXXrYfEFhpVBp?)FKQ*zWV3 zTG__$|EzNG$$1FLKw7S@kKZbq7cotl7RWI1p=UA76?hAQ&Co9AV46Q*#^7z6aypuz ziK zg0todFw|v+lAOWpC+Edt>FrxlVLpVEQ#C{AIEk-0Hgu?T{v;nXK!v2%pzwWG*49PNCFn6s&F!ymDyZL4Xo%UNzXvfTfWoNd<+?la~*uyx*?GM3Tkedp6Vne0m zmfGIn_x}pSfA;&oc6()yhbL8$7}0ij0NM{Ptyd}i{`g5w-_ucm`ygQ+OAr1%N?6piF z#`w$oYN-T%?|^dPShST4m?zV5lq*vkVEQ582Y!LZh|Wcdw6q4ckpT2P&_8$3&;(&h zvvZ?_d|w?APQZx;)@GB-w1tRp_8IuCaaW31B4y8~$#{ehWGYSB!{|2h@`ZArC(l=4 zhNjzB6{59|^Ex-SK7+s-(jv&ML6K+tMPZ7uz^3%g{YGiDpw_{yGaA^9pcV&3*6fJ7 zFnqe~J3t$9Xigr|qmsLzDrNhJtD$V~c0q-3?Bl07F3BAjC+OqBuYPk0Fvd7k$X_&E zP4{x&j9li9jgvr0&9QKi2--FLOH$y~O(NP?QyTroe5q_tXy5 z9Zo&F^xjy1UQrOU66*MFZ&ZlYKF;e3+21-iXedyb;)HJsc;~RuF4*5n16%j|0DNVX zUTcgMW0%53xCy2(Blj~D1x>;w8+y--U2P0;ol=qJl(XV~W1M8kpw-ztDkGMgR)x*q zMkBvac*31mglw>0AvF8=Sq=^PG=#^fc*LnRqn{bm{1^6K_(BlBI}IQ#AeJ3&gXKVH6 zRf_RH&YO$WXthvVKwEN#(8^PH1ak7&!-Te9qy-NV`!Fj%jQjQDsDrraiH$tfY&zox zq!Qe7Cp~l2fB$D7^7bIEy+Ye=h^`B%EM*x=yKYh`1hf_urnf;4IWl^8e*fPF$TFAX zpRsZkN%PyCf4ouKmZ>o1c#EbEMiK`+HQO!Y(UR+rT4)dxPe{wkl;NhdoLaA7o0 zN+*cGm{{_L(hatR87-?1;Y>>2mXk^nmHc<-fhu#?x=ER^k(H}Ro8RudB73631msDc zr1I{L4Nuw~pA9${GJ+wUVrfIS9UO?s#r0t~CNr_@QwSFlFN-z}Jq=c5g9?QHhj1~+ zalV5ca2sv@()}f7oE3)ePafrTSZ{=5_?|Pgi?kICz9=Q_kV7~fV4k&zNtxD{JZ6Qj z=S+Do00!`q8-(ym0Qbe_;@?y>5PeZawBNbLj8z|X(}$fd2Ml&<^)-$F0=Kqey^^Di zu({g^?cdE>Rv2zSd6dT*-=I~fK^FeeI#FPQOKS^=hURurpwuA5;b@V!y1l3*xt=f5 z{;%jM?9!si+HIc22`YWN;Yi#E7OZm^7ogN^mLZI+^w@SV^y@@bPIb3% zIuP&kUi`hfX0(YlU*WKJ#8)Br=cCJk&5xvAUrGLNcmDB}FWX-g@+Y3G=b((#x|oQl zcdw6oz6YD<{0D8L>OV35?_mtz;^hMZM~D|ayu6AJ#14BT28`}H zhqC2n+rFK;2GI!MNrFL2rxY(l(&E@+*cUG7H29=!kC`qwAhso~{ewu^rl6&AQ6u{H zRc)wMP|mjvRUy&hNnTqNcruwkd>aX0BLJ!B&Xr9~Z1gKWjE-ZA5p6rs<)VP191w(o zuT|3lXLnwa{GdY4#FIQVsIN#6;1neL0d->1k&=o?D!M^rEETXuIvA4%J6|?G zh2$UjqVT*MeCgwZ8W*ZsZE)&P*icDA_14jrEO&~kfC%qoT2@(Xd$RQ0jXoY17TLYLa{Q&OV4VVs{XaZLOyqcMmwtq0& zoT`w{Y$bYq)(ifF^e0F)90t58s*F_#XfOC&QWFGItVo>R?z|#K{;nlSg36?LMJ+J#~tk!O4X7Vd`t8>);O=1UCn*0pN?@0k_1B_j)}gS+C=}7$QX!F z5rdu}K&x(7v1oQ^$tCus3pOA$DH$=`M%Mn0&Vnj2z-KvCLMQia{<=+f&QP)lJz_3R z&Ni)ZR3f*mSg32P_1!v3-3&)V3vvB+<=o0P`rX<_!qF`y2Hr~cxf|-JV1fI&Ue%!zuyJyk-xD3VqJ! zi5^0tgpZ{1;*6Jbq=I=DZ9tkQFhH`tkHa|DIX|!(mQsfec-VS1Aap@|EH;-*s zBsXt&UX@d$N>ntji^J4FArY`>8ODV_K{Ll|^~`hM&`WS-2Un_M4$fE7h(SbILh}la zgE70r5tfuph_KD8@fhMkd@UZ6CR`LJ+_IPzDbm}WS7n=235Gt;=}^Y0_`p(=DV*wo zVV~FWz^Q2+Gq7*K94%1w6l<6sG=C%L{1AN>VaYXX)b_OF2u$Q%*o;oBT9Kw{#v`iYF_C7N0)M2?aLd&yVot!~!EJ3aE_i2?+b*tI9b~<(5LT zQxR!G#FTZUFyUqk;Ibg$E%%`jbT8@Q3sh=6i6A!8dfp8zCvB@Y!_2;cZqIcNOtE{ISDF68K17_bQo_l8xQ)?+9nmyesF08 zXwO_hjm?T?G!h$c10iKT5J!UyD0z&r>}#Hq*LUB>b{ zXdzO5DxYyi;HlX*yAhbV+K|~#L%NFDMMIbkD!-iv08#uuSe5S6;H#b^yie8&@s+k(JAlRk`&O^AFe#BF`8^#;Hi5{fONsxXcvKjWUTJQUpuKM#+HWR@(P> z<7ujtpnv)}halm-1x!TW@og`@fWNSa;9O@K31BBtWh%fXKjUkvv-i-UF?U$Ndmx4s zYl4t~N#HH4_yLlf|Mq_YQ@xJ_cCwb~?9Gqx7E1%sYqt@nFXTzM^NM(qDpm2HKF*0! zyw^xUh>Yh%v@*?eVkjq#6P^XdZ;c+gn7O*k)^Jh+)UljU5<7;SG9oyI1dW0P{}>c5Uo+>w!22v0uED>pSbhQ&v;u#q%ChO3BC%m*L+AE8fl&CF%j zCv$+3C#4{NJj_`!MG8q&VXXRp&R&-b2wK({qH&TgqNbRgu|BfuvWs*ZVfkWa{4~WK zf4H-LR3Yy8G;azCOC8h)M@Kt4@tKY+l^~5NI2N7YhCor^+(bKZd&u5wyTs1vnSoSW z2^|j%xKQ`gmZCLRzS95JIqAzz;3mqtjhcP2E%7{+tNQUC-&w6gvSTJQ@8c zjS679v%u5GI?7|jcN(%?Yp@qM)+Ynp=S~YqEL53DDWRv}{6}aXQfNM>_8DoyJB`}P zHM;{8^J;I$PD2FkBN%pU!fQBO<&O8NRP`UNhN`}W`BkdwKYf}L08I5grCUG)xS4)w z+ra_RjuK__G92ERKND@q4gge2&e%ogB3+inqStSusbA>W;LfWO5LO5>KFt#^^U*>M zG-Z&&%j0e1=+1Cp+}22?!Iwc)!WOM2ffLc0VJV-aW1QTSQveSfAJb-YfG`9xzGvHV zM8zYa?m(Fol{qKhTpJ;#u-oX$wZ>Yn%vN4 z2E&5>my-chgNxc>?GTniLZE2E41}UkZHFyL2@z5vy`+OjI0%@bLDHXLkLRUR!K$?X z@dQtPa+d$&1BMDg#b*z5$_x7e)yDcJ!!b$HCNt8d>3;>C|G|ci1akQ5a%+(0>bT|(%Q)~!>B_^;Uzc3fT;YDRFy`p&l?4DJa z2jGsF3gNVhLa#ym8$#C)^l2!sZ3=W*Rxb7 zUtDzY->W_pHmT;f8SOI?i&bMB5ZD12Hu&tGC`et9JWy;( z`XRmbtWZZ8gGQkP5!uCEerBv4zwY8&=%Iw?0`-Y)!^wcB;e}5W8C0I?2Nww?t+J)+ zsucNQL9K!!znunED)K*joRfSWGn}g>8;Cap zI!JF4v@re%^OE=koRC&)lgsT=Iii=M1tFKiwV=i25VxwxQTkbM_1 zVG7Z0Q`iN7Y!8_mw$}@F4nd2F(dO1ZHr0X9+ERUyxDOpUU(AB^LKiJX&>n`-||N3$=ge&7{mwi^SvB~qv8!`eL+sBnOx zIE4!l`$1JG*XbqNCXq-vw}R0A_kya|Dg+~+J;^ctNeW5(Tdk2yF&Zg1fJidd)D;ZV zWHGl~i1i}B&y>bX{Kysp+O%fl--8}-A+3FCHc)2NM}pzjFA zwA_e{eG*|VU^Wq6qPQVcX~wkwswNo*i3RKsjqR`|Z3kf=kbH>q;F4FgC(#&F|EK)T zKG{lnQ~??M?`G!}(iWdR%?lqD*&`wh3MnXA_Rw zba}xm@i=`3KiGE}ILK%~GD5RL{IG!l^3T|0%20wE0<#@JP;GFo%*=(asY~qW7r=>J zYNE4c)4yA(*56V3S0yT#H}Pf;;YE?v(mIdxOuAsQMV&icS&F`sz~N{eQ&p61hOnD- zU9=9?&-!z=W{RY;sRK+UWvy5{(SEpri@>C{q5haYJmh_w%v&zlLiqjbhj5`n!AeNV zyCa}VxbS&S>-?OChYg?zF2MDjx85D2cc!UYM0`C)`CUi=uTyh?LXZlh4O77-cA~Bp z(r9$Dd5>XD@=U*pT3jNs1hG!pS((peV z4_vPh4}AVGC+1k&tiOD_J+lQ~yAqQD$+qbG(PrZ~a}5j;wpL2#H;m|%VTzga*d`=N zE1XoUzJ^Fi>}hu#lJCBcIcDC6t!^0i5P7CO`?UmL7^h?ZM#PtT6dNio>kn=vF@Hz? zphCFt`J z5a5r{9ZTd#D~n3l)T_i?bMxg*=Z^%$K|rq_)INla#Csp~`q7t?N5HMb-0x`7Dr6Tv z&l@PU9Sld)VwbcY7$!iK$In?#W2s?Hq1S|DD$hK(TqmJ=i4@lp-D6@y@Hd&4AiFNs zW_G0CT=P$}uhRgE>lh>Br?N|R8*TMFTB-_Ry61V7`O~BN{t2C8r-P_&1$m94Rr?3& zIrTHg+elk>h_L3DCW5+=cIs?{rJlvG#)+NOuk#TFyuWgIlCU3YRne#5pYJ0&e8}J( zja;@>c7ZCAGJhnI^c7+JR4HZtd_89__Q|hF7=8)a0&mv{zg=0MR>6kQcXbyKKwV{1 z(`O-g+er0{*)dh&ncF@}mz?Y^Z3PR z;-@D$86gQOCK>fxYcr(G8+}P=axIgouiZ`ty#O*~cmCm(oF_l6*B@4?zy3V0_TNDJ zBx`xDQ@0nL?DQI$WHfFxFbX{#f^3cpTmlnUZW2i++GwwqYNLIKdWyGRla(t!w;w); zJ*uC?Iplzw96~B9grD#j>0x#fu@TE|p^Eg_*+LJ0{0u0!g{qA6b7w5XLhNeWIH)>Z z>ikEu*IObTp_PGfuY(E!1TPabRu#X`ojY&_wfusT*$BWiDj=i26u73&)@Q^ivL|c4|7>zL|3d_;ipvVpE9PUiU3W z-iSw|_@e#`fJXw3I%hj40e~2_ThwPE=zWUEfbvq!2K+>g?ls*!BvoZk&fDmm7jwQ( zp09}WtuEMqf~xCc0kWx%u1PBw$Kx-3U{nbGj$VP39id&@dVGYRg#>pOwuRmb2iJkh z%Z&uq-d4Xwga3`RKs_Adbj!o)U`_Zhc>zT{{a7+Et4K_pQ(y(e)ZHObUD2MskwfrL z2Pue>c^F)Y&wI(ta3@PyM~$Ywk4H=tr19%WWAWHH90Y<Vp_Ph3F;h5BT2iE#)JOQoRJ+q+EVGSr%SnW@rpF~DbMZ@ z!^23ic+tg``+;ivquMLR5tsByRjH`6J=UJC$o8nNs?Xr99Pt0`euMKvUZ@?$69?4r zttHc^i&H|I9?I!@P7h-1LOGV+MMyz;KAJ8l|^g<;K||%fSU9=?vrm>dn+>oXfGm{c3>v+ zsx-CkMjAynbSd!q^#hoV1InT*eGKT3+yfNtnb3DO8dkJO6JTVlkaSu27EBv0yMPtR ze6Rhq3Nqj6&Z!ckdy-@Fsbi$9MpiV6M-XJ8sa!i9XYUoAz6P3}iAng5r0tcmWVet5 zZ?|>6&Yc3P%L2yOPzbiy1qa51beRuy;iN^>naaBFa z@5rH6iK#uy!PIUF@GkCT09EtL*y#7;RK=g9c_($Djk_XkZjAf!kC)9!jk`*0?Ricfv)Nd0z_Hi@MKjP8pfzM> z;u}OeqfsMb%Gdp)pnM9)pDV>YK%EVG8JM7Aj z1H8WHKRdE$S6EjeR3wdKO&@{n$|0ei*0P)EHp<5b_ost{*;m?zWY zb_W!yG!WOx(b7<_bil3auXD(V2FLrbuw6ez8e{%Y*-ToB%^4U%Le)5rx>= z#+u=Nb_@w~@&d>DopUuQxW0A{2kJFC4h%`%6-)lq+8jvt%I%J-M7h@{qTD}zcT@)YsL(11%)+ z*P}bMSi2hug}6^3L9y|0;kn(xf_x8wZ5Mn!st}O_SX|7UXbi^}7fvx-Y(-+Lkgn3{ z={e?!eNX%G&PIK6vVOT5CZ=tkdSD%;ipg1R3`ITb=6WHd>2&@p!TWpcX+VLz+`~Lpng(dW5>(pcx(2Pp$WUFHiqn?k6J_J7H&Pzj4wBx; zjd0A8u(>0TVi1iz{SMPQm67iWr}Rpu^y3)9`Wv3{F!nD0{1H2IxdopWsfg# zLNsYA{wIR%Nu6NcQ`w|b*5Nqr{r)d~*D9Xy^T~C57K>7jwgKDZ{I|b-h{91Xz9uiC z3%qyn+ut;ZQ`3{*{{Anc-~M(0X%4=d4OiwTDI<`Dm0Q)SB&^pa64pOFD6UY#`pKh* zIR!5dy18obaVUl`o5$dlVGUT$hsoM$sLR|4ahR<@fb^PK&78|S05F8bG>}~Zl2Y~( zx12+R9`Ukib5*j{?12@Ktxmmnh4|Z}ye6j^C2HuZaSLL&1`%DN6fq?YC$Q~Mv2uN+ zZIG5=97T3|Vrwdbl0{|K5#!*xIZtsAFf);k6((Ilkz%*demn@SRJ*}3G$ckq{YJbu zIgXR-0A8Jga>a^k<7Cljxvg^$4@Th^;`ANHOHc(RCa2CB4ru&t z|Ms_2#$Y<-SEX3mQosGpk}r+n%nst;{&qJf@%PY;b|-O#y3tP_0SU z=mv%O7)<8>s;}VOc~?bT)Vmo-dvPON7kD&v1|LM{91Rb!O-7BBjka$k?R+;!afJ%Y zPafq>n%~}kqlpdze`?!gGBpmsDH|EqIu2kW`h7z)uOyKy3ckcj=F_Gfs9!wrqXg}& zMlbw~Nen^;fF`{b$0?*0=0%Hy(nE{2Tn>bu@z>cwg^*ycl>EbUpbC|epFGM_$C|)W zvx6I;AiaSUi*$awmI?A_kVxa>M=1cjWbSV`yrdsUf*LoB^d(sjVH4t60oD#G^*Y8k zA3cCVB5~fb8PTo8kc;hz^&dYHEK%m1fMf3YlYTYdOXI2J%Ikut(-eG{hlHGkfZXiIXm@pzvt3<__XIVUxNq z;fbMI1UWFbpJ9XT;Rrs8WnIOVdVg8p{QSw+ee)ezWQ7pXS{{C2MFJ3HK!cP}nf8;w zF$o7($V?%LlPO8!H&d=w#>N@CY3f<)n-Mc_oTA$y=}N>1EQXb|VWh}KB4iT6=&GfR z%-1Bqmq6Z}tx2zk@TBSs@o>nX=33)R^yL!#sN_oCO1QX?J#*(3skg6H2tlo_=OE2l z(KZ#Tbb$90$!Lcvzqjb3XQU)PKGul2lW{eZbP$WIe}t399YAkuUb$;A=*rs}jBF%%x=YF(Z_GIt0O ztP`yVL*sbqv|@Ak*YdU{ zzUWEH{rce`X@xS%PuBCA@?9IBcFbfj3K{-1VI>V&5Oxz9=Oo+S#-9mDf!V#V+uYWH zzD3y`G~Mua=w6@nQ_52TREI+j>JAR@9>r_@g01!1W-b^{VgRww#il39C1FmK-9XPC zemyKLwt=2KT#+NELQ!QlP)>4kY9d=h(hOP%!Dlt!J=NkC;iA!_^)MQBfo+B9+HA(G z#xp-1RAlsqF|)IZ>fENdOlTdh5$$a$tKda{vIvgNV0_o!S|5JGrYBSta0X3gG(xyS|8F893-m|4K zL#pVt%1S%jN?*Cq0mGeF<$$RWfn3i^xc%k;Bu-@?Yi^&U1w1;EKHexsyrUQ)`b|H$ zWQ4<#&2e`dVC_*w9tIWYRoMU)Vxa45IdlPZ#WZAvWQ8-p^r9ytg%Bd6lVW{UaMJY(ankiXA>xS|f@+l(1f(TRCIzLfF8iU7)VX!1D0vPs zb2p}vWuxODd?Z|UO!F3hHb-?T+azv2ret9bB*4MM&g!_?(?iAF=!H`5&G2y!o6 zKNBF{V?*-0NN8TkSOkgbst%VG5U+MfOoigrPuACS>RXTsS_5eblVKbKtf!)|S$*rp zBsL){nxHjAwub<~$5Ds&DW@%!&J7U|@4L2doG|dfJB)hBMC!rwhS52nfrmPbl?Z{~ zotN4x>y;@|+ns;7BYF%&^+|<_)L>INna*?|V*iHHOaGiqJ^F%MnWGJ6TgdVr=1cc1 zJF+3@IqA)2qLwGlmj0PM4!zpyo-^HZd7`r0q9T=Pw#5plOuKDSVaLz<)0`6|?US=n zAfi|~GD$HUY=XDp*8gSW* zOndw#gg59>DYT!mt!CIWwX<+^9b)k*u^bumOAZ*q2}(eD!7IWnYO_rz)o>Z&^#g?G zs3DcXL0Gukh+Y@l5|7iw=7(>I3iF!BYdJ8;Y=_At&>%WUB^*x17iX}hT23j|A!rV< z|3yFvowukjWBYRl21xE89*+CuzrhuvZEem8M+l*vskZU^KYwCl>>7RsEARJzRuupg zrsOWENQ#^Nw*pe!?%=2}sd=2oVwR?!N;vgHiEH;P()fk^m8^n*b-G8%9LuC>1s8Ke zU4L$)FwnrpmD@;4FnU^j*-fsMfze3yy}KBV!~)(%iMuQ%#H~<;jF88<0YC_ON==46 z+66?&acv`V^PW%X7`7&_lH20T(cK4hP-m}HcvuQmK{O!=_H04{#m0q-2f|ei>Ln)8 zV^QMkfRUs$-h*qFv*KH6U*FAARmi+}oHuL!9Su@p4KFGd4+%Xy#?Pbwj-Ph1vX_U4 zWb{cARg3>-wZ#4CCLwBc&S$(Nw^nv%KkG9g1JT%U0<)jSNq1Y)i3qjdN*KljSbrel@YZZS>zZmhaO z)&s4(iyOgzCP5k{7xiIG9pL9*Hr{B3{CuC8bNFi>5rmgzG6Wdw*DTkaF4Mxwo~qwN ztl2rhD#eX2ha7-(%OXL@JeUEwD6Q}pldyKqQhxv62fzO-{N$MIRgec= zJB)(ttB-0E_R&Rh4yXz++rp=9Ff1Y(_JV6(K+8hY9-MvPB!BzctBxVM?Y2a`+7s2T zKVCwA9Sr{*wpv%sM?XFK>Cf&3+eGQV|NQXL+EWbVHe$|&QXKBQDv;U3Km29$%_}X# zeYy6FA+2~A#$O%r5;tw40- zQdt(*3tGN|6=439tXP_sz-TD9YgntF%j){K0=~+h+oQi|asRF{PKIa)Z6>IisgHBm z4TjfI?QLlDS*JBG%i8Sskd97mwnEs7*004Y<7OFqWl1-oB>|XdGv()kQ~*Z@gz1vb znZEd4BxeVjdQVZl=bYj6yw*88mt35KGLbhiEOQq|k6d0W?2WCDx}OY(@d%%&z*a(H zP8!8!BdWL2rz`#|fJn)1+jUNX?;tFlE^CE2)04bG8Q@InjYdNh?GwTks$Elhva~Y~ zL+hQ^IeU@}K2jW~@}SIeD7wy37_$Xx;0I7eqYI72V?%5P)WPP~^>mDe4GJh->Y_*b z7sHZ!tRsiHk6mblMK+%6O4=fl&tI>8-A(-RutFs0$y$zz*Ty*9>A~|4hdYhK7l6A; zCvl@kQXrwH01Mh0^C0aKvi^%$<0%x_FwW35$lmO@d!N*IHPgn%P>@=7>M}gIG3A&p zn9$nuudhZSXUUycBpZ2HA=L9EmlZzID-w~;&7?n~Z%!LP25W(CrZuJI2gGYzmYR__ zk${`CQw>8(O9$@*Ym=6W_bDT_I8KpD>Ys;KvFqk76G0)|{Pj(7Mfu7b}cLH%#RqoQ3Xa=isg(+4bwJh{&@t~-~z6-Xt2P1@J!4R zo{YvC&7ccyl9I1pb_LYyIf{g?JT3VK%~LRp{PF-sajA}*d3n@qISJ)z^DR3 zt4k7y2q;|$hNALnD2o0|24II8Gi6eu{|T zfckJNA?Z7EP89-RPao#Qj}|FN=_ah7;X-8X78p)XyHotIN#tnf1dfW#ZxEd&qh&Hm zC(&oH>mAG2RF*J++ZGPIrdQk3dyq>V!~qSs;-F92Ni*cwDGfE78MM_$=WJZh@`+K+o&7*Tz^*+ZLK-D> zBXWRz(mw~iv23$}gdv!@n>4e>cut}TMY^c1iN|D+!5$)+GzN0I2mO#R!Y*TM3G5ac zH}EvPM9XTIs^&4H!%H49w-KE#w#yzrTM>A-O10~!dFTvifgKjxMo)iyi%;sQDg-T` zuIJ=GullOXQzV$yPNkGedW|T(0h6|ebf{@V%^gM=*y%2eNN?cevdOfAZ1B9^qxeUZ zZ8}X+=U|6`jV|6;CxO&fjYy}+^$=+~xZ%<*0_hwu{?Lv$TFxlVGSK=?531Wpb{E@W z&wqRmQM$t_L|C8Zjks_F6o=6lZH1Syt8<8iEBur+SxL33O{VgUP*r`y4Y^qd@wV^Wom#t4r{3@^k7CO{O^ zb+55p6Cmgyrl}-m=8nC3#~4pl^5W|gdGQ~fbXO@a{xol(@ic(wx(D8&N}4u}q#{9! zhj48E{x9u-h&&nm{x5_@Fh|JQE@@fO>4bP!hYXwj(T%$GZS}0Nm#D!|jbUOJ-cXoL zx$vnfxpDTy3doJ8r&5KW^3x|d4B$8YR8tT#x9;_-DPiV;yX#(Qrr(ZiR7E z>=1iAaia9Z3py{*XJbO5k20;{5dKPdSKlrckChNGj+=S<)2T2kWWh`aqz)UX1A|tF8d) zc(EkiJ#`o1=~hhZk)DFD{k5j=)RuPe3`}rI9QG%X~L+XlHH`D z1x1MQjkIk)plBso;34VR5r+ih6_U1wtSAt&@QkBK@ecZSoJkC@#+L~qKv_twb~Hjg zp!PA2d-az|sI#c#8B>w|I^Jys^w-nvQza_>G{?IUiAd3m;t)urxUv(n5x4>83$sQ~ zE*TrAMV||#nVi(`$U$uc+8QQ#rx1sYAwfKp*Zr>~I2)i#L=QK=O!{LeEW)8EAZ!L< zS1k;O!3B}hH5`}3nr+MOnC~IX?iQ{}VfLrba~AHzP4wH3ujLBPSt;hjYYNJ=RJ-U0 zibb-HX#$t|qhT5`oL}HZzY9GDiVNZDbOZrXL(7vK#3ZWVwJ&304Hi^>h9hYJ(=@(f zgzyH5XVG2N~qj7Dg2b)xm$vn^-0vTHJM?@|S;^Qyj zC-j8KsOZ!zxiP+rT70@ODr9p$eV)_mx+$@q1BybIGU-WiAqyt7C#KWnRG_8N#^*0A z6RNeXNy_Ht{-PHc7lx%;NH|f{GWCm}&>DQ6{ ztAA7_kO+JTp&Cpfj;tW4wQ*(Jxjyx|lNT>+Wj^VyxuQQeOC&X$QUdVt4XloeEvcsn zo8@%2)_yWQ@*B0>IIg|OUpXL70mbKI;32q2@hx?Q1&?{F9cvM7nI#f z*S%0o&z)CFek*wjlgok)F=2xqYV>t^*+uvNAWX)25vy+ z+^NApCC4q~T-b~?e@6#s)zeI1^gy|IB9hmQ%;9~g?k)1bRHyUmLpQv2DG?BxrhJZ3{vzr zpwbgw6Re((I3%1Ej2fGK!$rjl}nXr-* z#6Czyia)1|#|Ob@pr3c}uCKra}%zuDYbl@^XA2uUYVIX_jEM*jLwtDupe z9zGQUl+W^V(3v!XU^b(`x}_FWIo)ed#_Gm0-+-G=0%9LXstKouQL4xCK_><=>1Wge zZDMw+__h-ctRYv8hg5!Hs00$ZF1t%^qoaREIY@;-;qym1d0|?Qr(LBTwV;HWH%MQc zR+1*B;JK=scBZFKl+SlM_;&+1qA-LUH#7o#F!rZ@A!%qS9zaAT0YS}#Kvwvn)VJz# zAbW(kY69$+#RIDnr^lq%pRP(eQXwArJg*njpU1R|U8E%}PV;P9&8#cPri-N-zmu?Ke1gX0Ia)94H1F-o^|-%61FZcMsL zUHIpD{K1S3j_N!4VI%3#E|g+83iYf`bn2qLMcsi?Ex~qSD2O*)+aU5MT!Z~*dNXF< z*fi`g%(`?2@T4MbcPw@VwB4N(u2S3md7hx^elXVRXqLn_KtbD_?}U!!HUY+5qnY?R z)v`;SzyfWI)WuSh)@T~ck=(DRF#twnF;pMyUR^6{B0UK|DK~s?KfLWB=X((iK>4o# zexnw)4!IDTj1Bv(WFzo-1tkex64)NQYEq(j_`0t@4W{li+0`w4(rj zR&Tehb$uqlc0-t7S+QTBQ3m_YxpTs*bBZqe2{J~THnWO>M?FqpK%|Yp=~;4ayH<}q zlD6#>77(UHGa`wk240D5Km+S1LByrozHg<@en-n&A=LN$Nlr`8b`UYZhax_BxnGwz z1IQsAMke&YWTrZO<~{A)`W=6U>vMF^spUv9sM7UJx9(>2c zd)wi4eMnm&atws@%DTal=}HWz;sCuphGILV=`pvH)Vd{RxNNmD4myjW=j0Nm(* zF#3d%RAbKZ0XZ7e+JB8uy(MxG4~ba+I0;+lp^Y0ZLnW$8b$$IO?z|#leU+-~&+|0Z z-XP+m7EmelBRbjYBGwb5rfv*tQQFYILgMAp9Ljv^w;SUBrFWMBcT5CkbVIx3Fj5w! zBEfXK^QtU&g~;9Wyb;mU&f_Yq<5=h>reetQPGfFL65bH(*Hwj7e>2dW`h@3-kFW+# zf)Tzx5@TOy4CVc~Q{*`K5;2=@F97wyyKDrV?pZ*t&kqzR6`=miX+b){#^c{QT&Pk2 z@{AA(Du)IBT5e%hg6^f}pMQ>*@0ntn8*Bd5av6}$h)YF(WW8LGU=Jvvo zmRoJY6r4Kg(LtJtZYEKp5F!XbXwc{yBd-#^bjCCjo4Y9k^|D?S>8IPBS7awtmzg|8 zSbXtL$Oec_S`J^vnlx7*#dQWdNux9EIn|3 z_I9kMuU*@`T^RKAd5yv${tm4AgxV6@2d@XHMbr%cO`Tsjyy967Yg>^DyWM$3c1?9$ z2zE`5!UkSK45_WpXjR(Qc!8{Q)Em_~FebF-gHY=j4H~aOXLUZXne$W! zno{-=@Qch_Fbk(jLgy*ja&AquPxOY8l*x2Rd6J-5 zlZ;&FAV824loGP6A1L`a$j`Kg<`1-ly$RNqtokYY#6F-V8nvQ-vhCmQCqpb^up zc>J1fS(or{htZe@+4Nq=>^g&im?d6vW8FqGzu1YkwgNN|+gR1HK#(fS(fZJE=K$WZ z;_d^qhL9?rlpt;v81Rtl9g`o+1j}w5XF5*c79;y&?yY522dgl?54v_F$?qM6oK^(+ z3IF~%s-5ztvRmZ4$oe~o?x)Jc+n*whzZjW;`=O~e^@jR)gzlzDjM2MN*)UfZnzWhB z*}2rbIrPoWdFYB*uvD$XHoq~6zb@RalJ#8<>xu;Y?anI_e>|+Nd8AX^)hiYQyZj&CIJ>?sWg8<@${UzLzbrT(jplvY|$pereXVZB>0byMTo=Nt6UWf zj|=CO&7`6knHc}PmX}?;_X|X(`P*0hbh)<)m#;gZRtXjQUpWfQ5s=&lHqOn$Gj8 zpe?U||Km^B3bC#AJhJRf4QwBEo2{Nm$9B5SP@Cis9R;|x8c>F@TH1R9h}h?Mw(@vc zAk%*5i_%|5aNie6Dx<7GZR{xSvwTyy^7m1}cjjFo zvb3Hzhxd?@Y}}{tu^-2%!JutlL-9vei3(YNs(IOX>_lpq``ZYy87Ux4+Tb+-l6e-P z%KQnF&aMnmpAW%|K%%(?Dxx@-`abzY8-l}~+&=|1iQGQoY+!{+4McVn8y3+MH{d|K zz_?81𝔚^dJAU5+eGxsVWrFfBN`g4*qd+xP=N$&?dfOFjv174=!!wg=}_aeA>)` z5i#<0LMNDs&CKDnn2JcYUfUjo;m|s)Ci`#B{WJMldq^|_x%o6iI|_A1y%b1(kZ!<| z-li%ges)l%#Q(z!DHSU5qx+hZCX&X5glDde%C*zMdUZj(h_M$4mx(Jt`6PfrL>#Fk z5V4g3qC27z%xwtcCt3B(dS>nL*io;IJdUfXm96HI@_xOjb~k}wg|hWebH{+wHf*#q z8c`vO2DCqAE*-saJaH>DgeQZ$=N4cWqUtX_8jcZ_Go3n-PLV5v5oVbaqAB`iS$nKV zY#w{e#O8mzJyrgZNW z+J)k?va&`$V|F?zke30xh&D#cyi7;X^AS8}>jBEcJZFJx5}6c7cy6gSX-DaQVzo9+ zm7Oc2^MKYQxOouvPzK?5gmMvw0X>V#7KdGEd$6$LYo`2_6V!ezDfF+OuZ194?z|$9 zY=t80PjkmjpTs?zmWrw*IsCD3d=z4P4uV15SRJeEnq!iDP*%eJxqWdQoJVJjyfL=; zhv=Nx%f(pqEC)TAcVifREP1luN>2TbEVx2Q?8$nLwvU~ZI}D6db?^m^IacYXK=qa` zKBfRoL1gKiVH-f@NZbbjjN7?Yv51Fo7j%;HRyRg@ToBILkbc4aq=n39A0e@r+E+tE zFar+Vvb3UG36I~=8s$P^e{T2wqP?b3?f)VD@89vS|E6BAZM}T?w)Tjglv=&6-;(W) zIP+RPT6O>F4?&AWrOhGu2QuPTj}z;k{_x-aJrhiZ|L~yCk@)w2eBdvsx6n+YC4CR~QTkB6*^N)kZf9?|E28#4{UQAfUi?S? zDE?>R$oXvI{2wpv!~Pt$T31c|k$-kC7){~NHgaA?j852ZLWS%!vk<%aHW;Y3)_h4Z zWf0U~9sH6p8y-Gu+C3gzb}#?-p1lEL2!)t8A3v&bjYvS#vNDM&G|h(` zA8uMh{T0$=R;YQ?j#V-oYumC+Zc@Fzn5(t$HO&`o2ms*amY_lz#p?)6{zLdH(W8E} zxmcn_7*M2Ftdt|#uNzQR z3r?18 z{yfgdW=3FHd`?&z;4hxAT?FDdH4K+Rx_n{MfHZRw@Tnx{>31-*l#yGJe+WxJc1rzC zFC^3xGGiyNSH&lB{O+RGn|#dbCSCWUj^o8)v4Bsc=$@RxosEx1^>p3H8Z zjyz*(`bLf_tS>lmksl)A?6Nd7D-2Ej6V6@ohRr~waQ$ci4xJU6Y^|`fu-%n=MyqRp zw&qJL@XlAj8<981JKQ|}=J(4&-%C0X^CawRng|1<#(M20BUJ`r+%#g+V!NDNtobWk zoU)i854utFRv&dlhr`1;)8o?!noPCPSr#=5UbChUyoM-+$kXA$3)qrZ#R9t-9f`P5>SB+Yla@QA7OD~t;;ub(46F4Dac%;zwzd#y#c0@j7AE6N-t#Bh z|BkN2?ZiYv9j!mS#vF{*p3W5iIiU#wEN<4oR9VtDvzwf%Fr}Zy4kLRVkAT0ZGkuJH zbyg;xq3G@(Qxwg#ohe_P70%>~J(krqIaYXywcXj{_D15(?~Zqg|A6~tg^=o-!gB>z zQ!5-bp@2ZP6wR+m{=V1%(`!W+Nv}1-a#M*WjPV)Ir4JF!+o8*J-U$>F_Z#Z29G2PI zW?=QGqTcG8|CrS^8(WhlhVw0t({EjG8@F6Q|Hk{Z=P4>m-O2X7urD+co0Ae{6E)q% z+x~`{qNXi1Q0?xk(KX{xWnFDoF|(EMFR-jqy@;7&PFi2RVIx>Y`u)r-T;riNRme(2 zY6cfbq7MHkY9>Yg{?)dWrQn!MAVBx zmX6h2Y-4lpN(H_qKE-8G@4~a%UO&F3<4jUsBtApP7c;grAGTg|Ld(zr1aXhzwu6~PqLn^bl{%-fnV0Od)VBCE7pt=2^`94hWO503nBrif$6XY$V zIw|UqTahWEp00uhC+$tUidd{=tul#g-+c!zfz{A>or*@ErlQfFASH$QOveaKydJcp zmpyW#87q3HRA`&8g-jj)Yt2>&(aRMfYiGbHvx zSQk&V-ukO+PDN0v{~oOXeAMtrldPOEPmxM29<;H+vO*)iLYM_%H=9eZ*iSyE%-DC` zK{NuF$9CbC)*22j1pV#lHJWq?SKVv7!MGd4&|~B$7`}UL1Ar2NLxU9q<^~ou;$qKj zbJkkC~5n(>_Q@hBwA z)LoOkGy{uTVYxqfb*)iLxWuL2<)in;;LYl5yTHHWewk0>A{+_ z@D(b#1*zG!o=Fb853tgTBq(QBWPEhzAFTtGmZy;6i8E&9z5oqR`Z{JNbh2E~thhq~ znN}+q&S7xDvv2c)8l&R)vZ(zO`kPnRY|_PBVmXDLuHW$9L`pBYzvcb1p;tXPpkWgh zb_X0DfmyX&H4iq~`-_7}e^JdYnz01kg=sBzOWXD*njZU_UKN+WZ3H1=FwB%?wydLC z?8^IW#N9FPENT+p^e3#YDcBM$F@$mt*Kc2M5pxUaFSuX(8|pL)9{uLKxc9Ni;SJe{ znp5)~iZ25i^>UyNwOQGP^tq-NPI(7p(jipPX2`gEg;420PT!kkNW)b7p+1|!llvxS z7f19|LZ}dW5+E?yw-=emi}0TuM33$)L~Gbmm~Db3Cbkj_c05B7g3P^w^1$z~R*J=C3 zjHlyc0U1+d-xwGlq73d3!2sp4tjkHFW3>H}=b7~~{|;-`xb~v{?py!t)wM_cB}>fl zE*`Sq`rev~?*MMhT5bm5j6vzsWl{PnW+$fIXM%q6d9taESWFH3C9)P=4VExYP4KT222^ylp55Yx zRpLRcu9exsE-|yaeDvN#ytx(J9R5c4%hry9qQ|S;a?^QC4$+5YcOrE7rJ2duK_M{aLgG&py@zD-u@-%g? z1Q*-v=XoWHAeqY}Pq>MvOM6xJoW1Rm`k`Ed z1D)!!m**TMg(<-W|K!!RJzKzKPSB$CLeJRmdT&rgSKZ&~e%a4yd=>>ms3k)JxIy#y zw7#=-xHSbgZjHmm=96Cf0;)+0P1d-U*_mIE!I_MhCcr!lNZa5&ZuPGsgdIgDo~4zf zpDqC=eH*hSjKrx-Kr#^m=?pBq5N@yb`Jm4L*`ofd*k8T67L+aD5+f@7jQz&WdDxCE<13p?$OehU!U+=%)RqZV8CjW?zUrrwCj>W##o zDhd#=k|#I?^H?x$pxJyS6zEbcj@TWr@xTyU2{yG}37I4db#vyC{^v~4#~RQ#yHzNmgU za`uct?_-I45wpBE$=EU3eAHBC3Z<2SjdHj9vmn~A+zJ$yiHun)B0n!A=6z4L$-1+5T!l~8^ z`IgD0Q;ceR+{_oijbeI!_?)%p1!S!Wu2%yqH2Q3Cfk=rH#JB#4Sqq(fBO?yj_cZG} z2#4V(YL7?wZD+xa9&vX&s2>6Kwn1SXB5N1#($k*6oIVcPNzpX6%u`xnE1f*xrRF-h z?5@_q8`F%D-rYX(_rG7p+Q6m|F=aOnBthMLb#i#vtT$g>FtI6zsisSq6J=M_PGV+> z;WiBe^oR2$vaNc1`ie?CrqY*N31|vq;)2@IYW1Jrr(U_>-xrTPVXxZs` zh|*~~?VbV6qu=x1nk?4& zj?3-8!Tqv2rb7U6M&0JF7}&m(ZxlB-7>1csRm?7{K@zpgYDZKM<5^TzH?-*!c7HR9 zMWC#MU6kNuUA3Sx6V;Uc1rV+4YqyMVcUVF-^_&0p)io?zy(M|y_gf5aC*D%?c8kCA zep%G%kd0q|pJ*IS^90IEzg>riLg8^Uyog(x`rT?aI%H?5F#f{Ey6HDU7R}yRn|kAn z12Ut@KH7}!AowIAXmZw-R)j0w@4@d*hA?A}qIg7+|7*kN7LS3+KBo9zxVnaE zEuuZ{LjX5p8}?~(Mf&X2Hf;Km{X6~b`=q`z!(hQXOgpVIbCU}G)vIeIZvB?n$(=oN zZwKCZzG)x$x8E-_VvKpvr2QsrQm=x|**YI0<1L7Rz3(gW5{vBH@Rgiuynl(4G|{V_ z=G8SQ|7e#vpF3OxZx`OImhJ_A`Ta7Z_dzFu|9n1-fft%7zoZ#|T=pMU-f#bHJiNwk4KVT9mE$kTH^#uT(@TpN*}6*1HOUTh>ei8kwcMdP2Iza z`rVjrBEmIB3ffil_=y^JGVWa=9cwRb^)S9cbf(5udi6xc#^jD-DAFIpF&86d(z4gk z6~}YZxc4spM4Nf5TQmT?{c~Ggi*u{K#D&=fKYE3{;dwhPy9@mt@0TeH4Im#;Sn0q+ zsLoM8i4%CAF1TEnW;qneUz>%Ng8y6t-U;)}_yPEDsnn}|atg7_{2W99?ZphwDOw8Y zpuyI;$zKML*%E+`=m`d0)O`EnxtF4B_>(+Df_+et*ONN`A50bZ%T$JetLjZgJD45vuZd-zqnTha&14L z#qGfY;nGeWeU+lNHu*Z=_()dQ)U>oqY^};w@OI;^ve2&ax4B>D2P8G7Sk&ABnP=d% z&E0?`CEPK!sT`CfyIPbgF_*Ketw_AoBpMM2rNNy$o}nO!hJ8(0NjX&vkA}k zdz+B?Izl_9FdwpLBmwlEOnLOM>l@?hWW#5jgCBw+&vTdfcgR)?8jS_kyT(GIW~UPY}rS!Q*a+Sf4a zusY0Z2hiWpHgu+ip$$@k<|aP{{Z>vQc6742ouU~pG=yXFewx1Z z{#y2aX`MR(b)g<_`gH)y`bTE6fqB@Z^|ah3dp&8P%!(hRsO_YR@P{*Ln!c`)KBj!! z&ru@BJM~J|Uk4uqhsaybB5qTx7+v%st*%wtk}t8#;?LOIiMPJ*w@v(K-!B_wD|bSA z=MYd1`w$hs2!}k4FZpqRJ!v3ePD1A|rz89lgg?9d1zxf*aw1;wDg+T3x*2X6d9ij9 zcGO<(oaTh#IAe#_31{EU%#QKz3O!SI=)lRBp|{FZ>nP}7^6l*n=3h4hq?e1@WuXVN zx<;uLUSgAX_q@HGOlC{q5lGI^$F>1tCugdBRG%91ndLvbtoYf1OfHUL%&t z_}@R(5N=LEn0bz)-G47gMv^ruJ}AD#Gg@8SWEGa#<_}vWZ)e_=CmVC=9`kR!Uj~Y7 z5ULTV0Yv;y(gymZKjjjmV2((y6P_Ux)@ot`Q8)Ir1s|Q#=eFTOD5mW9MnP}LXKV{G&h&ESYuV*#g?k5>Bt-3SM~P4>FYY%5|y8A)Nay#jy#>YCIoK+Rh|PQTl|1y!};{*&&P zU7dC*6buc%WWeLAehXTLkc(dP5Y4UQs1vp6s7@u%>tajd>lCLghD@ezQ1dsyx)$8P zaZ7&bpEEI)ik)IW+#HgWa2}^7BgnlC8ByTAsB>Y9qr#WnZZJAedL(I$Pl5AoZ|@>f zyu)X-x;E*a^RguE4p+h(lQ-Kt**N~T_sd2FR*aQ`S#R#qNqzg}A!O+7Uidi@fU?l@ zxyvLen_V{vBDWE#4X@+Dxus*)7|N&;dK4DTF>npP*>^Alq;997j1DMyXZ3`=Yg9|# zGw7ta_+IuAxBfF%*O+XfmIRF7{7n7M_6AdCef{0;m!0f|O@ze}{;-)jt+3t;uRt(% zHMLY+uAyvbp|Q-q0~DTjkX_k3OqPIuk80|_F!()fYn$;?`ecm4(H~P5tXihqX<1DL zZpR&QaiZG-?-|MH%5D=h(M-(HeY;Ije52UYSY4a4rCnlEl`Vj`7H@mSwu`^d{j$1) zkVyr~&X7tbXk@G*B{quR8=!`iz*UI8ihD_Z%s6=uN(H|>dxf*D>>~>OO{;4{TA?Kt^sP_VZ)k5A%a_t$f4>X>`C)ed zek4o+$^}hd?EW{jIU~+$kalweQREz*j;Z(5rr%Su_fT5Ct{(^e^IGHba)2x{q%QFr z>25Pva8|5pR@_4l$G)1g6KIRzew4WM@`Kvz!Xi_J|J>EJ3)XIlS$yq5`fcsa;7%9S z-~N7Cjaor^QNS2p0KNDsXr5ljLK?#0qPaLo+oaDcq-&D@Ih9@pZI?67A-j_XA==TSeV@$yxGDx0**i!qzN?T=Vp1IQlTGBe1oDe!px7WNjh zU@iSc_iLXk2l$joG)*zm@ykYovXJ!n*>^p;3>){-K(aq4k=I)^-6DIBuP63w^oVm7d=n2k29xXuhcd&S8-y^rc*uh}aL(ioSfymzVp zZjfA#NR}B4{(v z=_)vn@lk!(uM8WvHBi2e@zPvrboo3_Mag#da(MHQ7z}T-(T24p0-}=-CfSi-0N=*h zY|xHqr=+;>A{6r z;k*@h`c@a72@UO=3OV3wnfZBySQ0&9XJ)_mNG`I9i zd|H_&?u}-~@T%G<{?qQ4iJs5|e62=ipzCwzz@Yw&;z2~%sp)9>CW-W5dg1&SwHc$* zje*|M&v7B-C2DavC^V>2IBqxU^`cHls`Z9Qv9Qa^KII$!n$k5?>KzjbaKrUqQ3Jn> zhcFc;^r0D>u*oOxOPE2pDi6K+_7f>Iu_IhN=8P2$-irQ(t7{HgtR>cP`$y{cw6}qR zOX_cQzl=j|d}bctoenwXKO^w~2vZQ|93~E9SZp4%STj>Jr`|DR1HwzQkFssg+#ZKL z=)>(5I_HVNm?o6X7TRzSgbT}r1Hh&&9C>9QE=>O0{?64klCMM5*XGloot7&(8LuG*d+;EjYQYaYn#m@^g8nzA0}6-G1pz`1O|g_jI&EquIO?H#mmZ~ zxRcqZJX^@ow_tL5`lW1nU9;ea2u!i4AIa|bEq~#uVePHk>G<)LAdLUzDxKC#Q3hdFo^00#m#2&LB;>> z)ionq$|Yu0@L_u!@b*P58p^daJD3U43oEBncsfp@KWlX@!!6Mg z)A)u*>Nm1Chi_X-|Ni@B2rq*{ulbfn9f_qgmZSHvK1?n(0!r)|j-D1p`%d#LXS#8A z0kg_Oger$M+In8X0mA^~NX#T=1FP(_AjI&|>8>9{SM&i<`ca*wOR?w5ez8rf9X_Se ziw0rJB7m>lMoB%m;L;O5GD-&*B!kN?y3ua(|6RS-;qg@SmA#g$C5rqIs6DT z7*+A9UbpvDA^CVShxdH0#)YD7o`Pm>UZD_WyTI7Gy8)78g}qFnKXr9&gH>A+pcQ?% zew%xvD74W2HuuX~)YKd__j+-A)I3R}HMt8E>YDxVda>J>y3l5CaG~o4)gX?z*M3p! zrSh+}lkEcsq4AFyTd|MCh3NCm9^NwAy4O~WMg&=0@A#Ylpw+bpw<=3a;%m>)?^(a~ z3$C2M(f!)|A2^H8;^toyxCPS zq_1ky{LneYM-d#E@>5YCv$UvVeyD>{#2I5J|4_R|4)GPd$5 zZ9m)R+QXp#Io0n^*`fN6doMPvb^>nE>ZbZv={0a=hCGMDL0YkgxVl!yZdhV+-*-*C z-FX8}H|m{lJpW1e%fgw)i5}K**lv#@e8O&tnvLOkv(frB0SIVfdW-Fwi!d&C)t}6Q z8lw>a1~WJ4O+ZJ1Jm;6j(G-m!#k-9bYCw26)Ck2{#KOJ^oKSrck|TXNySae7Za=8~ za(Nl_fFq7DRA=UI!#w*R+=v65=*^{XYH-R+uz-741e$Y zWpQX)8UFVUeYsXwBhrKgUNVWpO0vh{$DR;!Ad6h%n_Q0L%q}-OQT<|g5t(JVOl(-g z0V?c#7xiYxzfFAs49a~gyU9*qVv%nK!Iu>Bo70LcuK`)|A*`<1O+D0QUUKTP!q48@ zhqt}L`@-Mqewo@a0|UsJcZpn6*zYKg`XdFfC*jC(Mi)k}lXuWRGXNj-2%B4TK# z4N1pgJ)&b_6n#W{PNq8(r7ADyo0w^BiiiE78)@2&y|5AB=QIbveoAXB{D)RpLQ#8B z5!HLHXl0Xw?LBaaiif0y9?a?*-PduEik`o>8*hSTcaFc~{j$W3_&V&TWnc!84u^#w{tbFk_{5+bm!ql|vi@2L$R=Wrc zWoA~k9$frSTwTMmhqla*-TJBeeeEr!;G+5)-7f=?RQ))@LgJ+Pl6jlyfYQ*Db+aH;^nbxo12G+O()+C3pPjp+h1MdY9`O0F{4yh z*G4g&L6UD3X5XthL}Zu(AFqQJlcj(g8Q}&GeIadW@|8>toNsA{wi6A;>M%NHvS?Ae zD)UFLu1QTE*fPVKT$TG+{Z{w3bEm89Z-2im=`Hf#3`>jRAre46dXR?Feu!I|*DDnV zG)NeJ?GmQNE-Sm-CeEbOvT0*Atwyg2ZWqDQ7+P|b+S7zlJ_{kHUqpi}vM=Ynl@hIP zk^+iH0Ez)Z;So;mt#tzyHLS_U6nO@#Yg*ZYE^$~zAGNmzZ)|t64g78ImywAO(P^r4 zs0pvTHAETJ&4tI@Xlz=<)UfMRNSjs7d1cG;@04X zYiP2sXjYac1V78sDGy6yLMyDLW{AbT3;XG8nyewJ-zn1&^^K2cb&bwed5Oi{?c??a zNX|Vzu3Gcmjc^!sbg6k?`+t9 zKcV1;&fW!VHu@Bd?Ke!!_|lHA7`7JmQURl5 zM$PoeBwJ{)}#lzEIqs12kvdaTi3Vj1pn^)WkHbsTFo~-hSY6C zKcw5lpEy|q`NBBTT`X7RJllL5uVeitqht`K*!$6dDN9zqo^eb1XyY=rob`o0^`E>B zuE757l`YZmsRLx4CPmGrz#qE0hGPr1BpUqYr|NgNw~aC@?C*BJY~>`q(h0n>2fJ-^lFbXQ3GO z1?@+h`TAafyZtC(390aVnxoZ@`9pec5~y9SYIFT1B*&sQR_p<-u9>-&USem}E`v8C zZvsG<6bGt_W#RA_G@(V-E{zhyEvMzF`?8P_sk z7e;2$#2m{b1wP#OGh=+I9)11nOcVbnqa%v6pI=9Ge|-pl&1z{}Ym^TArBicw4Tw_I zM$0^>)wNW6$jd^>idMlJlQ+@3+(-WI_sdjY4KD(?SM5(wiit9>oBLw*FE+~tk>zzt z`#(`6Slb-;g9`>WiO8YwQncb>IC5C}Lu!mQ!j{*|k1RBnluxj(Jp(ue1Fu(yv27Sw;p{W6nX(EVZ8f=!q6 zqt6!)k$5Fa3rscL^(s5n^Tp;PO~0AL%s2YXSR;N{(jlqM%+A7QElrp^EfU-uM((-feN$wQWhIp`i2aU$Mg6% z*ci3#s5jDx01Z-`9ceR@*XhGC9H*y2ld@IB5MFm!yfA#?H_U8?6-@Cu%kr?gO^4zi z5mqwtMk4NoDOZJ_#_C#{t?Uv@tLEW*OYxRfe6RR>-7l+ifOzu_RA7qONjr+`Z=+#C zeavAi4NBAFVU)wVntnryu4Hd;L7@riWc|BntXb_FMfD+6N`#&Qg^{yV5&I??P;Z`Y zi+Tb?s|Le9Xd`VDjtYzasjF*OZrzp`*4H1W-`L(r?q*f}eeajSKm~QYkA3N!Z-l@T zVe>sC(uH8XE<#^M7tt>weVO3$g`%@IQw%zeCtu%IW2V3i8`n{HL)ebPsL$J4rXxLz zhDhX-ac}iTH>AXcx#Ak@yuZGW!$|_Vv=0ICJ2$&tJh8(*?XSEIj|cCA#!Y+0{a%`` z(*r_l?q$+5JX&E1>DxcI)wMP~h-FB}T|8fJHr^b6_#X0K>3&&q%Dz1a@IjOk7LAL~ zI@E$ut9zjT$rnak@_sdPrti1nT%#Q@sPQVa*^2-x49kwM^Em(R>&eo!;?Lud!F`%I?HOZD>#fyC~Iod}WvIR!Iee?oCS3oUX zX19~jAu0hA$Kl%gNULZBSFFZ;X{njqmny?B>VW%CF;WhIRkCko&J(ae?$infDUniu zKQEz|Rfoo9SDhr-W0@jC_L<-MBv#q>mig4O&)FNq#MYlDQT2as~hD)ioI*FJrjQZk!5(>J0Tfqd-UJb*f`L zV`xS;PQRwz{~dN_^#+%XrVedC)$XDf8`lM{Ns5_)plUW%(!;Ey!iHDuL9MRcX~CD6 z-5o!5Z#~`&E8IT*!|s=<9z^E>y{8fmLHCrB_{ACeTc_9P(l;A9MF)3L%`UPr5{(vo zOpLQ;f~&3p)wD=bZT;G5S1^xBm(>3LXzl19kQnLFm^x#V>34vZY2V4r@-QB>&yf12 z&etnGHT+xu$kjElY@L=E+1DSV-^$)@?ruf>z3-Qq(1;dxn|q7~rW|lWNiB@kQ)*dM zyE(1@u{b#NZ+YP|FB;LNR7RwlMJlGLc@{Vxhe5A?8eN37t#&7LKupT^%HcD+Z-$G~ z_cY5vyf`%3EK1W3;XLk2wE;i&va7KZ@ujtB z-vC+yn>2-1F1GO0>!lYuSd&ceI3S~R<-yrnpNi_~=a_j})t#F#MC0^=MG7!* zs5jd|x2Vm1>z}i_cJ}o?xNm%=fW`ri2f|7sPeTSZg)j0IOzMZPW0v4*OUTq+VwLua zVk1NaYxM$z~W{ zHdE3^a~l+1W0DLHTTEOR_-Kkh^SMEK~C_Xb*x zsG&%)r`-+CQ}JTi<4H1P${&16#g6ofB4}Q$QdabjU11wamu{IIm3Yv8(|b$$);0Hc zxL;<2HpigH$k*L)Fi^tMbU1n)wSFaD?I!YY@b;_)-6*~{Zut2wup}C;89H8kR80A&u)3C& zJ?~{MD!ZnNHSvbyP46qa#lPWxSz2Q-fX8HVB0KI%aKq@rc2+{c{z!F2wcLUwIzcS zGUbBi$*)Nme!ykSDlGjL@2Q022oQ z6Z3hOFbTe>S(XZ7;%fBd^yKBz4xBh?Wgd}1G!Q(p-;_BEA|3Ah*wUTWMx05cX*#T zLob%s6@DnIYl?1#m!x%f{OG;8c$>V7&EoHRzl=~6f^C8CjySTmDYPVq-RR2wh@0pV z6R4ZgGm+2Q)Id1RknZ%YxMup2N7k7N5$peF@9mZwN0uzzSHacnZkuA+vF@?H$QkP~ z>Ob|DDp8f0${RF9LL^co!6FGMsi!*6aGvaZ?h$YjDherrRBjAxE_UrC6##+}vEUKz z%a@BZIpW7Y!7le%7EXrJ69h~=fl+KCaWw@OaWJqZh#NR96X&GFC3(MeQaZ`s|5f0%yUr zrqli~L(z4r89rmp@F_L|jCqAcBod?o6pl>LM*yH$2a0}h=*zpMU=t?{IJ}^nE%}dr zsoPI|Rk!rN`sJ~pQs`R!7AiUV)rp|+kA!Whe#D@OdH^qls<^biy?MxLj8G=M44}Vi zE(~1;;^OMrdWqQsng!plMC_bil%|)Cw2IB2I@7p^20O0=NG2V)TyAr^ea*J*e|mjk>87*s@6mxEKGzT7ClJzNCCXu;#;1lf}a;5dd@IpqK5PwDib$RKeY6nCiW zG3nt7KMD?$GA?DcXR|2(@;I6=`1Tx1N&eMdJRkJIB*}gAK@z#3$02!iQdEb8b2iXO zM|3LZoagPxYH{#Oym9}-wu;v&o&oV&~twNWFyU{S*~~Q1^p<@83Zr%n^U29 zlA#AfK|gB&ErbB?D$B0t-P~j}Ym9SQ5@jg=XH1RZ&>(M1@xMi=Zv9X9-xrH=^-tR_ zPwPRm-UPRq92<$!WUzJahEj!ukC&p;!W%(>y#7_NY>oJYw)U|9Q?+(g0;Y8w!L$Ly zBqzngUeS{4N!C4HaVTQ`t|&Eyhn<(6Kv+!rF^xswPrD$GFK;&(`!{+<&S&w z0Mr*#;vg)+)Hb7QtA7PU;=0e*icVJ5HDEVy7PKxuo_#uRf9izaK7RW1{9`BlG#V$a z3W)=n`}bOhk2U?^VPRag}RJ;|w_O~P@Pl)5? zW%7^?qtqQGDbNb5XpT%S=gIM$?XUa(8--gVviWfGCZmMPdcOSu?fLSGVtXm~uNJPX zIwAWwT&-A=ZuToqqUfOc@hjOBKhfP~o7luE!slLCdDf zn)6C5U*hOtxM;07VMURp6HA`C;|LX?ClAJjlOOp={*0cV*$p1WIB#-^q=WA`V6#YS z0NtVb5bZ% zp4jt>^n@VRJzr5MwJ3878SGVdw^sT^^+lJ?ObNQ=?`6Yi>()?EW#;TI&076uxCy1? z4jXz^C;k>s(^hAY&Rw3;K{|pRx_%|-`Z4ug`Z^^gOMjzLwH0hQ_}L8F70p`nY<9j&LJlc;oM}Qm(}*zM z^7U`Wob@Z;uJ}M0O=KwX=G!G~=R~PJ-=jG*uw=Z53uB)%&n!V_Pnq5#114MQUNS{e zjNkyLEA=rJk5<}QSfIZ13ZSbf8w#hihhwrk5a|XNt(&@N`!NE68pD%%kJUv}@<&!0 zulj<^&3*~GwjXf=kU{QOEggGuv!E+=s-CH&1$d8L-U2ldygsd1Axo@Xd)9h+i8EY2 z7k3F|VP;=(Y9bw0-1LUs!f{KJAY6A8is#s}(vRnj%-)RCbwYW9J)0;%r45w>&;%?QVLt;Ep>T2(IO1r zEa?#kN5Ngpm~$LT)LVFCpGkX2gX4INUqb#`0F#8Gp`_$7F1M}u>XIGgG$Wn{ebQMp zL3@bwqKoVfQNP5MXuXK%(T^23Vr8GN-@n3duQlwLvl6SO6aDtVwRLZ>+S#1jt(yg@ zvj{x~fz)kA20Nw(p$FnsIMRj1^(exz#GSjGnOZ1HQu5#`r8Y9{*X|f!xdUGKQ^>SiGW z)hy913BP13=H+m77x$znY3R0QIX8Jd9F|t5*jg*hiTqP^M`HP<)Frq}r-L-b*CBNh z$07)k*TgE|L|uTcV+xa(Wv-m{0ZeEX4u?_jwX#=y%b#CgurQw`VA*_tZTTB`b~5dD z+h(EaHanBZ5l8*-A|0pQY#2@PoloY2a9NLgJ({dNE(Y4K)V}fDU`la}llkJe;B!26 z#X7vH`AuPQ9^J*m;7|%zxdK@QWYHrzHd%RquYia>*SPP?a%c?>%3l3Q+|f$U+vn1t zs%5G;6!k@xOlk?Vs*kW?!63HtwP25E7TlES{;V6kYbe?IAyb?e?QuF@jNsa@M_b-h zycgC~^TnNuo|?grQFlRQ9Gv%4IeVd3seX+04nQ6+#Za-R+8ssYX(Y7>@WbEcu8iwN zif7|OG{V4Y^%R=rgJ71t`FGbBTJj@GpjB;fZOt33b};F7)20E*`5mCRb3~GAQ0df* zeRU{s2nKSc*H?&3FE0!HWkvpjE2Ek_CDzg)z74QULXLI51?M{dLky9#a&VY@J!09* zkkjV&gks)KvoY00^t14H`5R%q%Jxh9fIFhnkkl7y@`FmCw)5dPz!(a*St@q7X2DNf zX(44~!ap+NO!&jYo5du~%LuIXMdU35X_*%fuuL(DB73k|!}fUg1oux6vBJ^isTimn zUPsxKq&HU};3geG=yd{nII1{w87euvje8FnOnM z9BRXXp~|Lt^YvgiYZj>X@v99qID9zkMv@O1QZ;*|(#z3AQn1!<4fEE4w8mqBqd{A1 zDPm2@j)Y|)W)fXwq>?A73>7dX%sqVObRdFFm0Kk0)sTQ9sg>F@XU=4MNR!re<_+8_ zpGH_f!Nslz7{MLxnz`f4(o;BgdEl@%9o#iiV( z`-m;v`D%+w>liF%Nq5$A1>v;P_}|Q<^bx$dF6*toL~(b3qS5joCl8bYHs7&?d>hU9 zb=tI(^+}8o$IC2xa2>+p)&^jzp=%p(~oXD zd>to}98;Z}yv3;07r;8nCHQ@}Sup4rW-BinyH&$bIDoo8ld4s0Lc^2A-8AV5v!uv9 zem(QruK4hht}M0>DPi3YW0!^L1ygq7d0kFnIvyo0RXafNhz)b@hAZruUDpJE(o;pL z{FCkRCFyEVr1dLOyaAq5CaNAZch9Ic8uf*rpWG4vRvBXhgkkGl)`C5zS-3mw^%mjB zVL$v6b2puHV4RabhKpwe^)~`-f z+mp1?X*7P&Z}z6?j~vBSkET^dt-iRF=`R80KV@bNZiebztT?-Cvp|gNAPprDA;FRE zRH6zn$sS|L>K0$^)f5qu8CpPVPp!pkB)wZ~WPS}9a-t(Bf`7vzluw||w9xGu67zQ% zKti0>RG?P{H(j(u`m0#f5QZvwE%FGYu%+O>oyT)fqroS4#j1Ril=_Q zh=e)~i|0m?CV;lNjXiKL@f2@?-R%abzW9|HFM;p7&4b~{fV=eqvRgI_nxtyYu}ax2 z{1VTG@gls+2)Tp@(LBO~gaa1kFIr%hiN-vp$joC{vBH{DJ2JUHf&*I?fMpI-w4}?Th}_S0KM#EP|uR-cfX7Nl2G|G8j-U! zokpf|I2)OQZB20{``)KRPjz+M)Xx2qh0_R^bO9n$j_D)ut z-L_fi#;%|&A*`eGh;r%(U$<_k3K;H@e$fjxKX5&S&6;XU5F(j`H!-|&QN)(OEDRd5FeP-7^(#UMiiur52l8{mO0HJ`F|NTfd-(#Cpio;P5>3RHIBw3m#4}=Gw+( zLt)h^ErFglxs26z$(2T=zL1n@EkWcjoDf5h0rSmk$S!Rb_*nWd<$OuAd@!6dSWdI^ zbPlQL6{cU_+&O~fAnTjEv=58UwugCjB?+}I#YMzf+=s}Hu`U==t<}O|RlBx=Mk?p5qBy*oaQwNFNEDtnb)w*IMo-FZ(-iP z+f*2$46Zv_Om^F50rfcU5iQ-5h|MhF0P553E%oEo`e$C?&l)-&uIx}X=^m;ec5j|< zR^bdyuv>j|{1&ijkLN@w{c#XYr(~o=V|KIW$a9RJ9URqNAusjnB^oOI=K7+|Mz72a zTBFT|*jBxPY0G)H8#W6@-{K)<$fzD7@=TL~=9b|75Dbh%c7^K?c-j^XXriRPt*?$N zmo4>|9HJyoD;dUpXa(^1Bv_KfMUSc$FVUVTfKFzxf;Cr!^QJd#em8XPQXelVbV4*# z+CTz=05tSM)JGgy`5(xuY}qy+k@_M`=C$mACuNCuJ?;h+L*DkQ#qQZG@Li=XgcH=s z5gw8|+|>e}D`DQ8hag~F8`-d{DE?>cg`AEcSd@dEDe8OPUDcJtPGUb40pw4J9bDG zviU3O8Y%(5Y%b16*Zd`W+%aN&rn0#xgLc0^jr(fc3gBAyNNC|JHHA#hMxk^DQ59WK z;^+X1Ob)3q)c(DyvC1D*aXjjas^w^w*{Wr$3S({%F;MMrY1qx1g+I}|y@!+LR6yHK zkIiG_5yjluO!RHmp4eRD+}H0Ki^2_5aizBq9GA@^AvnklzF>5-U1H;Y$#fW` z46oZRDZ68{KuR@aS0tztlO+K&9J%@7;`M|M*G&1KRCAa8-rj17hvDKw(0;)dZBHiT zNkcwUHwzERPjle+e&~ur@Xt~D7!3=+b%YdpsLy#(B38qA=iH6W=t}Hh%PO&(3`c!2 zs4^@i5Zq#54F(3RcU}?pz-Hl2wbO>k^bR?}9f>$SSQ6h!m(;@bMp0TR5=;|dpIrT^ zhn4j9 zq1C#}hGZ#9NNt&}zjn`Q_3rF~V10`IaRybh7u+qZ8amrYh+F^-@uKO=xLu)JlDkM> z6=l@2jGy!9-0%9;9x-*v(LWrq6?ImvJw*;Rl7?5v5T{iSe_IbxeSugQ;Ib27q2bmu zVOTPt?r;U!&6@>O7lf98UHk|yGpAsNolpNB9{!aPR}ee)vqCMMPyh1o{~D~c$GR1Z zlX;R#Ue%cjv&G1P?*M1MWmLR99LD3;o;ZdE@qDg&R(M`W%*X|A)ER@La}>uj7tuR_ zzyeZx;-EQ8(beEF)kxEoN2b2G%xAg8M84lV7=R3&`CgU()*SWR@BKU@pL1Ag_=dQeq*}k!WDXpOFtsweMrIkktpe>_e9%={3kz|Yj3dK`D9Jmrn6B7Hq z;XR|iAnL0F+Pe+4q%u7K_?tnT_F$t0=arqTTMuY`@vMua1kAs9h76I0@b_AW_M~QE zU%dnel+BPFM3j=jdoGo%^v+h!?+mXNQ!0!0)}#M#my3`jH?zwFe|hfCg>YlJiQ*FS zdgMh3p91y|AkIi{70ZgTAp_i+b;$~zdh`huMG zPy+GynFzy|A^CmQls%_eI0gz4^)4a;EdV0C`8N^7>MUUB#FW<%u?@*el$|VG0nrBV z#MQN@V7(dz%C(oT>UB78Zi>ndxa_D`*Nb}mt?b85GI+#n?Ioz)e6$TIhRAnY7xt)Tff9=6A_*nU=AqQyhSf_xP$FKg z!6)mlag0z7nq8^IHq%602m;eI>kon}RYZ=YK;m8`T8)%mp)4Y1s<5$?&7tq3f}}H! z9^)Zm7ME!>s0xzmKf1nPVFF8_QaX&bv<)inGOzZSrXh&4i{!ZPV@a+e3Z@arSkYPw z4+zG+#q?L(|p@y%>v3rn#`y1A__T2h@to%$K>gWCSk0h z_5GA+093K{rpC9_^_;gP;5Lo=lJK`u;N6M!uSrH&J{ElVzBrrG9Xx>-e3!%c2SI*p z*l}g|Vt(h%e`SqvvYfCoxLB*VjIS+n!$zeUw%ardE98!nmd2ta0=GnoE^5=}{JE)xmKe0IDRyqS6y%(scWFl64?8!X z!v8qPeF|$`czd6&dlXF{)KB*x2YIc+t8^Bxq`npasL$~9E`^_PwPHeLtE$)# zIbFf`LvoL_)mc2C%@;BiM*8`fIQo1q>jKHPm^&x{bv z1Q+xxj?fWzI!$9oL;Eq*-*%Nl8+T>ewm*L7p-6bBU6U0RI+fFn{guu-Uz)@>> z6%gI)asQ*}j*Y3dVj%C}7w}yyN5QV+*PJElb&OkQ3s3b%m0v=Jr^(cLPOe3v94;Io zdJ$x+dMw;%9_#zoa0{{Iz43onk*>C%tRj0^*&A>Z*xJ?2dexm`$nl!G6oO+?L_|3i z5o32|ZCxbn##M9UghGw7eOdedTV5EvyfE+=y5Pk_j3laUcyJNnS&Cnh(0#gs>zaGp zXlCS5#%WhSdnUU6TbT!WdcSCtU$B8$uo>}J%JEuA>-v!lWUWQX1}8sDf973voTj`5 z&$DjTH_i6LSl`!=TN@>>pMT5=?QZ#XuZMHyJ^s z4T)Fw$D>z|>&y0mwhyemd4Qu}Bt%{(9?%`&BNN`y*l8n=~5Mt8m-$^h=$>~0PWWQN;cxW4S+%|f_lx?>r`qah4KVI;Jv z9hBTBBweqy+Ic(4AM*|6+%bv;sA{jZ#RPuti)0q(k|W2Z(}mnnIMJwSmaZruMSll3 z^}@lE64BX1^dv~6eB~SxVO1<{^jqr-KAEi&+-)|NwyX_PyP9CTbF%=`NweEjN=&kQ z11U4(#)c_VVh^cl9#1y3<~8_(P-_V~N<`kX-XNP2ENxSd(vu80YpcI5N}ioEmiNuI*{^%f{M8h> z5X@q`M?oM*(Zzy3U>DQrTx)`hE+wd~ z(cBqu?Uk^*^=kKS)=R+^I!wdQFd1SjsLRf70cME!UET`CeGF8^wR+o_t z7h>f?QbkuUEU@#b?S+WfOr-H{PziGHiX*DSI|h>M%L^|E?cQ&J@2!G&FN?>KR1@oA zU{A&Ac=5abfwnynrEF7=dhB!D_GHg2nWg7hQaOWW2V+%Vl)D8`g7ckDfuYHe`VK3| z9?~qxE}|b|pPw*B07gg0xYceb;$+(^TU23e=ocsSC>oPdDsaC_TL=rEOWuN1=0>CbLCw z9swn1S}|P|We={8_gXg+Kkv%p3ZNE-o%K&|VigbiKGDT%q7wPmzf%K&ojBV+{p42M zxopp?0@WVB+3?gCr9bC1+E3nY#0@WoyZ2Zx_LOF!vXhJm%!=-%yi}>m9QC>gp$Hxe zW$!4Lr#{{Fc6IeS!DFFZo?fTtVK3l!&1O{MOw{2bb!Nss^#vTYGn|QC-*k02#>r_q zaDan5Mq%ns%|wikQ$q@CrWpda~}ND9K{0 z7PDXm>5&VZ&^<-n%c2Mca#&TERs8++#gdDRa^81B)3L(IqCL{W&!JjJ!Lcs zKT;O;Av|k!u|x3>yrAK15KZE6GEETPT5rHsS_j%XaIMvB&EmfJWYlIA@S;zGUZBra zz#JoQK(G@_bdG~Yzq(|fcm^@b8LrRq14U{%1b!O!-QV}p#Y6fixT~({T=AFJ7gBBx zOF*^rp*Ktz?%up6?9yhz=@xFE)S(kwrB%SF7Jv^_s;*MMbVw@68%xpo>nKGh*Fp4K zy)LhI9?zqe*|1L&M8%S4#~E;HLoU#Qu!&c+#LTs`AIK0y(*?zmUwCOeO+|l#qR^e! zPBIIm0LrX#+Vi$UQ(xe{l>qd!ku^*hoK~06J1+%$c(VY>7Cj|XjJhBc?19*W&u{>- zW*F`PgK>;T-xb_8Il8TNey(1Eo`<_6@Rv9W0C#DU}VGs z9%r1=CQ&lQ0BRNm_75wi&WrVU%bWX5>Sc4$wlq!4tK=~icS~Voc?>x6Co5b|DYcBh zV?d#Wzx0b$@ecjMxgd|}YE?kx58Uvl))!1nSqYf78e?1S2B6Jm+pcdGT5iQa#dxqG zrV-C3px&AxDM#r16`Q0{Z(4PHE9hzCd14>6R=D`{iEpPpPF62)a++okTcE%ZIUB9= z#a%}f7)q=XwFw$j#>PfJtiFh_@hSs`HG1oy+F~{^Y(A}ak7fblBpyPyz8<8#m$b9I zDpafe6isPTkXU`qpsaZPir3>vU(Ni~-5iw7lMEs_TfEnk(p=XhMS_khUMLLq>{8Ik zu^A+>Bbps$M5nJJ;d@uLFFPBH`r^w^bcxleFy00YgIB@*TYVwejhlrzXcj%uhC z=^9ZXrG&5#U%le~TQ)T9#KL}zn@PDE?&}~zC>sx-KiMR+SuE)aLb~crP>M!ap*cW>W}#sptRX}4~r%Rf(uxRRyo4)GGeS1*8Q=v3z@ZSy%do}%v(@IeH3JwD9c>>BE($?8;>7j?0xQ8~~vf;wID zdS~kh)eBeVRQ%%oPz0B&Nu)ao( z`wBx+UntpVmjTKez4w$Dhzx$4EhM|XSs?s_TDBbXl_wWE4#Pz>dA-z`qb<Zgmy5#wzv%(Z<}Yku34bi&v@U@!Y@f5Y><%=b||AT507~9r%xREqgT>3wHZ<> z+K$u@8h#*$q$i0e{9kaC4Tr5Osfly~&F5)K!U}t`*J3sHcs6U=A%)xxob^$xrQ7!h zlw#2@4(D6W)kzHZaS4CTepIBe8=|q(rwknUVt#5v*LYxWlDnH*Xe7mC~=l*mFk@qH@qU_k^#< zr)wxBhOCN^x<$NzLFWPf9oyJt{jtLN%xhTWXghOGa|IKHqeU6e8M9mdAO|WMRyAK6 z{P@mBn+&O~W`oMcvud|!79hSQ?5_Huh{lD7_xDLx_GN7FBpnIu4|aH3neP?b zWmjOY+!eCXb5|_WFEUZJ3Lw#;YkleDu=3bb0V~o8d>7gyMM%3hL^6jICQfx;%R|L8 zPsq=DO4H!rRjE4Qb0j~0eRw?r8(n^AB2hjGBy8>{7In$e_};YeRzrl#wXx_j67 zUwZ5IjjjVyno|0s0NWoRh5o>~i+Jw$sHfjuQHU1yLk0y21syytE_Guka*|sQMSU^o zr?&)!I~;#Qih-}%`mp;n3vq48;_Uvesonl(>uSH4a(IR8TiQnsZavWCf*1K`b_LKw z0*kF&7bbV~pX}%6|M|ag0P-LJqz_dHok2w+Fm{Pf6(y+kE6#CGw87yxh+Pt|6R!#` zvInAhl$a1fYe=?LH;Gk$bA1sd(^mqgT@Je8!0`0OWndRJ3qZX7;h#co&AQmFxR+4U zm3`zGs91N@n(6T+AdudOdE!UnL_GNOR5W|niAoLmdFIXE#aU_?%aC~H4 z;pJ)9ocT&KQWrtaJ~d?tFYA}*@qWhnizIrtY5W{d@8Xy&#nb2^y{ihZsv}Zgc*)$B zz>7a`J=lg0gVjzKhuywe;QN||WK4utlsk|h8o5s0L3p=#1#Q|2WJxQ*ln4uvOl8MK>i&%r9HBYw9fZR8B+3YdvUsc$rrA46~b4pB;ArLgZxE<|4uc| zI&%z!%T9V>R5<%+rxXzA?X&(RgrXn45=%1j&O8^HlI*`H-{M@NQ{JcUT3-OT?J5Or zIYu=G)lW-5zSB$?m<)qkFCx2Tv+#NI?5cTv!3ZUPWl4`bTf`c~lL$`ZOW}{ zP6CEDCxyq+r^0w{gkLy>xL)lR}OqPvuy0( z<-V2husFzDP|CI0WfBvTld3yhEtPhLsdIKH;|FwB-2zq{n);%VDK4>azi>hfN`}sj z*OJ|$S!ldSrsBnj=A_}WM|-Ar7qTy;R~`PC45gt&OWVx4-K|-$I>R#bTc=qPp3EfGeVJ+)?q6QE?jp*3huyR_Q#Mog zm7~|{b%9-nBcnxHAbvlggF* z183xjaPQhBUuj6P<+1bOc^7X~W> z>8{q2-ML|ajYlMYeR`N>&#FILsw-Ky;S`|8jnP z^{>Lwx{RL!ye?`*ac@#^jfi)5l@nrkE?`#s1!|>}4tl!rq!=WGV$U^hKk!ze@Wcpz zC^0ou(x}`7WC>|kPbSkRRvVW3g3`@$2`slBZUc$IZ}$ttp3y8IiW{6w=(s=2&>cjR zR){1JON3IAr;~;4@``JA1?NFhiRVDfu5eP`U!UkH+si>dfJ*LyF ztMNOrR~HC3rMOO3qTcRNB<#3HD}pX?=MJ<`OM6V-wS(1J#s-^n>2JZGy=bBB+V|04 z8E3>!bRYXd%S6aIy%Hn8llo3RNPfy;S3>JsC|z$pGW7+apWPB1{^Ch7G#Ntkl^48m zN!cBmg-1D%^}|0z9Eu?KY%bM;mlm?L= zoOnwYFJkFyDVg+4c*%6qTg&H2z^CPv1n(%C6Sa>OePlUkqfRtQrdUX=O^OUiTJ=m> zad7I3OPS{qQ0`=~4JZb+_gX0Sq-H@NzHo(T(ImXFiRgq5vva8L?n5<;gCC+hGo(r35M10NeIczoA&foPdV2 zr_}X**2sAPyKg0oVE-Ghjy`HC1N~l7_RbeOywo7? zYP=QJ!xN^J7sC++rTCXzRd@{ur3hw`c1B?eT>;o;2*CIhT#T`k92KlHLLcvn+P(HuZepTMl$1!kZZw|*bW*bk!l^kCwGAi*M z^q6uB{?fT6^cUYT#TgYH#?(#6bdnO4#Lj{I=^vMIG9*pvFdC(vZth5cx8Ss;noU(P zx#_6X7n48d8U1{q|62qf<$wNsv<)JLy!Tis_LOD;@idKx;VJvjq{xZLU3#yck_*x6 zytP6LL8uQ}rcLP(pSs;4Da1IM3Z_zfAg(Bnvg2$VlV5cd%||X35w{n+ zcsc5ti+OKI9W066U{3SPb|WrWX;A75MmN1BJKgsgeM5}la_co?w`>*&J7fbs!tiTP zsck=i|0foVbVc2?9`@5N(N@ncWoy0*g=1AF7ChRb+u0l0aW#j&J6!ktIc>34Rvgm|l(!fl77zG%!xtZc`c zw|Q+;4F?9TUpM{s>}Fw5c5jP>?8l`{MoV^$umZKz$8kJHfLFZ0>vyLnmEo_#cCp=D zUKRYQ!WrcM_c)Adap3($cNibHi-$%@E^&QPB;5!P@HT6G$-P0Y&>Nh<))};h%K9?f z{M`Cti`gqNa~lq=Ep!9f7ISPjYZhwGVrm|ui*_R}#xy)gW~7VA^7dsvSnkVO*;H8^ zax#0xfW5r-MMTqItIRxirdvt(!n_e#N!fEkEYj*##qFi*p3wEQyNayK=rOtzIU0OA z&Z&Bnj_x4uiH=hb_e-~OtaC@BQeT|;=`O)q#Su4{80L1oEbIZz!r?9EfmD63OHFD# zwu%Hxi0H3YPjI@$t|=Lnt|{b$u{cW~DJ#ss^J1j35j!GO5;Y8Ld4|$_W&oWqZF_{4z+%=!amC zOmJMz9&&u3&z4XNVhBiEdl88%0FfGW@rct~0gCiW$*tN)wR(;s8^r@3IaCR@GhxHF zt|L!VPhhd*LTml$Jf7M_*#oUkW!^yLfvPVwWy(vix%;s<3>iY-Wd+$|ngvw?Xrd<_ z-nK7~!}d5DE}q3`?hjL{!10gs2&wq>K(*F{09|W>>9hv7_j&rIaD$M-huJ<8FQ_pg z9f`KM9nB>Mv_lF@+@<3#SUruwT_+}QWKK4j{^J-p*vVQcGG6z@irS00MyN&FF34mxz=?|u12Pg z7Im4dhoswozP!TQP2`^>4>O)%7iUQ?BLX4zi!_oI>H4)4S-%(6zBzlzaquh~Px6vr zLcD>TqknVa1n0ybJXF8I!s`?8#}xaWtZvXUIP}#xH`T|ezSwj#USc@kX$}lVhRxlr zA-i|8AbJeZ1nQiubK$X5x7fIEVeZ)!UE!Wokic{DTA{ud*P2)KF%DL)6=JJcVs-K4 z=S%j!@uSpaT}eW5xRcKP!|!565CjG8e$K1$j1RQ&y$w|sE}-lSb*3sBcQG9G#ih=2 z2`rbR?gI1q&~7yJ1{4F~_G`oL**x&|qBPXKOq=LDaXAS_NMQXboj$C$ix!C&-Rxo9 z`YaK@zt#G4DVIuZqizweEl5Br~NHQf3n>e^)i-^uUpy2_NK5>Irc#aAHfK(yt~Yw_x9gkUvx2rCAMpu z(Y3{IklNV_uv<3^SG?)`o>4M@*%V$aBr?|v_oHD#sZp1IYHB@ zjELCtAJwSh7#E>M&w1`_oIVZXUjG%bZ*9%N&9x4!b#TL$Z+4Gc6sIT|k~2Fj1_xB1 zV8f`_DEtcul|FduXGu@;#)bpt-6$kKyFiAE4{qsIsr>8f3qPI45(8Luh;98Fuy!%~ zcGqU1stsEV(mf!=d#rYE&nTPvQShf;l(|(46p-HF zQMvQqINk8QlpS?#Anl;(!4V}tFiPd>)+bu6Bry@vA4#dFYmTYH?fLdZYO*_0 z&WVOrue7%1_|+H7KTl}-C(V=r(eS?20<_yT3)jL2dy1iVC*k#B=hL5;2jNZn6is`x z&`Eo&_on)`A4s0i#<>oNfVPj;dG6*(vpGl838FukdEMX=1QZmlL@mUk`fiwoy>(Fj z9L;zMbaADU`+J#LtK^_2;kuMAFYE=M|o9vOs80wA|&)VYOLh34izd=!9E(>2hH0n zsh&qrcY1hTlpmuEm^9&D)uQSf{>=I!#LZL5_HxSswqt+?4$T6}fe^{g*(e_V zU=P-lGoBK{h|g(J##cfI*4oj>IH|eT!90~1-5R|`o&gScnu3FvKu?5TQnr4g9B~|p zXiQZYRQM6~g@Q~*2@u{gjeXgG>`Q+d&InXdW0n0%sf zJQp=H>eD&5{7F2XQmIBOzNEM#j(*p+WEpU#@duyJaW5M7KuMc`S?sRsjN)uaZV%h6 z_H>*Md%>6AFZ{Jg4kINzfK@#SPJZvWZ{1Hv?7>E(D%zHJ*l;}Si@AKZ%7B-*squ^& zfDQK>EML30S?E9cvvYX+@iN^1)H(a}(#r`u`{cOOv*Z;Ur^>_LlmR^=*EDspaAjnA zf&HA8P=Nn_IztqORveQ9W(l> zjYoZ<>qfc^xbHs~1|7rQhHJ-e&@AkoE(R?a`Tg+ZNr;zupe=0QljB3m4{5aVu(#gu z=8c2AaiEQTgdmsRU#-bo0N@#xSBkCmt+JZ-ftR=MlY;MDbcDYp_*3-bt+*-YF@h4o z*UItm`~jQ(==wtJEG~oB_9Jbp-{7TwQ|DETPE~3mql*0~T<^!vjs5qbG_HTowbe>JZv#dA*{l z`^aj)t;R56j!F)0Zx~5i#0G@wvuO8f79h?ONni5LJ~_|wFfWO;4@;@BV*-4 z-u8tHDWWtNMzT5poeft?@k%~x0Uh%0b3JlBn>;s@G&CJgnb}ld_~YVR?>%t zch^J9Hx9J1GCiEyyc~2^ER55gchO?lRCo_NJOqYETJ)-u6g-cfpHryA;Dz1*8||CI z2>EGB`icfUn1eQ%^^(frrfNUAzBqDoS7MDeA6i@N2A=I_+wR#c0HNSBj>02qeWBe< zy7M#+@fHrln?d|aHBN0LF2!LWje_I!e_IiFtOok6G3DkRR#p?u-21EDCs><%D$u0l z3fu~&&h=Rm+9bkD_Gmksr#LKAaUB*+nd_~TAfxZjOl4hBuDpswvCp|Mantv+AzyZ~ z?qU@MzfH4oTW42;^GIymciYq&OfTxq0fKjW;0OVmjDZ?!rEgDY*1P34xrdhpc~17q;}~v6cMxJd`R{%@;VaBAKF_spm*o{7#Nc)Z z7;wA9b^>r5g>&gud4+XJSA?iVuN1ytd2!@|4-{3J&5!|A%8K$NwaK+? z9v(=vO}ft*`RB#R(<+b1;iCZ5OgWkZav&;GR;rFtecu2w<0Y?vcbx@;l7V$sE6MKM zERf!$!+sR*B~xr(cy)EuIwE$C*XI?@`1N32UiP7w9b}sVe6H82=TCkdKgtH29XiYy zt<}uQc-BP<4TaI-J^-E=7WO2PqAvbfU>WfAh z!ZJ8+Jj4csCB$|)|91Ok;p|qbl%R=3xx$mZLstuBkE8>P;aW`5=!2_+;#viVIGsj! z?4?&-ve$7mY|RtuF^DRec$HqbXA4JKdAb-tdZN$~)npS=IS^}@xlJ^?VXc=k?%(;; z1Bu;xO7jI9Y$@>}JL;-%+hjEA3oV_^5=*!H@iuT6_I9*R?8ePP-KQa0uT!#MImb!% z-X!L**?oZV=1XN$ctMV%PHS^L5?3z!usr3iUI(`fDvVS^L90J=xdXB)U@l9nYF#^t5RZ z%{V5GCJ{h8Nk>%L8^@f{NZ8><0N&aOI14Uh(5zB|g))EzjM6tTd)`ZwOG#u-7Rlwy z-Y%U?{_hyqoJyma>G+QLW!$7HP&XZ>`T|qtx&)TH9c}}SVe`G$jyHWnkVu5A(5$ZBqwk0;xBO0XM8&`aHMS#%f(4I$kG3uVLLy|r?B7yWs&Eai4klc34 zV4vlpFOpI85O}4Sti>O3+E#^MrJrA449X;yV6gi5+UhsF?P&h(#?68iB1Bp=A^b{q z`t?EEBsejDA8}?%J+D_br?mcHs+(A+yg9^xzKF(aoD;V8IQPChCJ2{xof#sfB94e@ zQQWD#5r#g&kH7E7PnCt_^LuYL3iX8-lUahUw~nSQdBc*dkawJO zdtkFL)=7Fj#8Eg0-lPPb$&$!u-TlT1hrI#l*?S}ijj8z{l98FMb&zjVyokkPN!_(H;yuz;1cLo1t-a7>r@<>D% z9tk%-lFl@Yf8g0cd`VeDoMa{Fupdn+P3_EHCwhbhJDxo(U?S|L{Vn{U){l)wslFJ@ z=eh)l?=}YpBt!0QSCZYoSwOvI>(;uFlOLhvRjn~sBr|lLl=VvW6IdqgD zm-IxQ(U5`+-9=vAPc`MzmQ}|@i)j+|)s7ahE1_)bt76m3xNz!gAr$YB+#NjeQ&v%Y zTsf1%p0?r0)EAyVCkxnb-e`mk8V0?0UK#elhG7v+Fb?V>7xp)qc-hh~{itg#=Su}} zJq6^hSC=_>5s%UrRmS9LKuaI3v;IvEdW0O1S#zjz{uW{G2#37tY0f>N(vPbzqI9ZC ztlg%AYYW?u^yW#mOPhs>bMYWhr3(Xf4J;G}*Ksn&so2#zT|WVZf}n80GOD1$(jUoW zVy#m`0jk8k+e>-}Wu9|iFhul1xN&)}6f*MKRDY#@a)G`)hVE!NSRz zkKTMh>I)>9+Y%h@YS0ZNhOb?$54&r#peJ#vBogj8q7`5NO{+Jz!?DXzTE8Q7tv&?p zz&F#Q8L1YO-Xe_x4BSgKX9*QJ`Rc%J4inLFCj_Ti6#O5d^$GdvfiT6392M=2i{9Z{ zk%pa=6JHdZrSar}gsArK4rd91o>M{_d?UIQPgU*KhNDnlh`D(z0a@jtwe@ePdY5%z zk7*Xn_99OTzs+E3PNxK}WAg zjUQ10ws#tO!-zq#+A^{GGz)9z$t*&5WgJsy7OOyF(JcIu;<*p0SVGPrg6(%HYL@E} zr!50%8Cc5?!6!H`-BiJ$h{G4LFOu%g(Z9wGDbEk=`LdauIZL*eO~)9)vuH7lU8uVg z#p4IGn9ju6DaP&~;sr58arR~!KN44$@2=Cq^bX{Sr5jhS^jv*t>I*>|ZZqu}aCO7eeOuJm$?6O6**00lm!!puV7Ub6H~gwi#lBfWc{3tHAEu zEO6asq6!gq*weAQ9M9t)b0TS>Ry+6QVqe)^=51RPQHE0K3Q2w(Pe=on4eAGi-_$74 z6dtiXVu^5=sIGglB$qr}ocGcZZRT7e>>pS7&oH+1Z@L2&eBHB=)Nj0>pR$y`fu<}Zl7NOc!YBJ$aS++HH3Jy)pPPkxa2p_ZH!&H#jlNCyz;Jx&s zJFrf8oS=JUHC(&xMn9;&c#>%;fzw-t(w444VY9ij>zlRoh0NrLSW4gxJD(0(7dOZ4 z%dj(oAW|B7wzgD^l*T#ZU&{M9(HCJ}IeiLEm|tN-t}0mg^qvAb_99_gNwQFGf^q_K z-eEe#e-R7;!R`c1gDFRrP{Y9hUk+hyg8w(2mC2P+}pQ>vPq z1M^m6aGbZb z#1OdRr?9ga6G?bgWqR2h0v8p|t+~SWJn0YSPWe{Pf8y-dp!wA-&U$I^KMwXUFAKIa zr~Ye?ba^OlkGu$k`|~#I+ew z=#o9@J5GW@$dI~&6=XMU7A`+0oZtrGbu|7?GP?0h!u@IbL=EROQ$Najdupu%ZH*Ci ztqWjtcm56J%7Yx*h=8No>QV$wQrXF}L#=$6tNIh_iyc23CFptksM@+UFuXQx*0NbE zoV>vwyu2N$Pna~4n(c_gKF3>gwbNnR7f$~ArJi5+!BL3ag)YfEzc9EUaVDE8-z)et zR~URuhgq&GnD0_^y@ul6(b4dHZxtsO-Gl7@KCk6ibsM<#@2@XTWXQ@6Z9f}QTk(dC z_nBvVPP34Od>7lcC#ewB$0PV@tbCA8iVE!=`PHupd3|!lfmYgy@}KPNWb3BMF-2s! z`+FTO_J88X_+-qfJrNGl@2;e*B4IA-GFJslg`ZSkc=&lKF(Va5*4DCN<}LGSS2YU( zQXzYRTip2u+Z<%bz#75S?S==Jm+MVQoAXQFDxeh!lzM`>nr51Hxh%2hXostX2@cpq zInN=Lbp~lqB{M~Eb3($(3!yCU5AKQX^#q3xc>{G@;LCU(R2FLAbPVf@H@6B(;P$Jh#878geaFRT4{H|0&%n1h z<+Q$t`szoh$_+!w^;~g0u8n>zo4eG|rN7bTxTyc)bJC09Y9VyD2>zhTojMuY`-7C@Qzyo^(mHXWYHdX+jTYs66vlqp?JYQ;^;| z`a)KU#*_edR|9U?Fd*(=Y1mC01~{gqLwqey3L}Grx(aySXM<$2-lWoeMTcv(eQ-x6 zKul6J4*r#+*ijtKIjTAOl?%WjfPj*o=U!Qa?(guE$k%{(f+Fio?cbh&?=l*>uM(nb z2T}69y8g2)$_+-~|7>(^FrK#H4Mp!a@%F4{f$MyjQOofdI^r|MONa``XyB#uP>PGZ zrYEA!*f*wV)k(A927VSHF$j<`1;7gs(L2LTl*cF>E~5J()Wf!Ue!u%_#5jBA2&@U=4qFsZ3})0uAfZ2UC}IvNXa(X z>v4z-e|VI3vsp-BSu$PL?(q&5<7kw0saZe|I#0;gnfcr*bPe3DXwVm0yjkTZ;6?rHLq)mujwfLWKf~&$N@5CE^ca0&# zoRtA)=c8>PFxYIqAnYE^0#fJj&&%WR!X-i53(2626ip>VtK9ix{F=QhG;udjv6(_19T|9-n-b zA9+FnY09s-bpc<8CFN6il{j%pN_L8eApdaLD+)-*6A&`bR>B@tho#1Ndjscp7ej79 zF%0f`h zxI9BKLa7SUs39q_`|PLgn1xg)j)+fk?)qNUW>xwX^+k-0U74M-Mw<+zVL@VgzdrDNhh|uUDOn*gv!Z&{UUd=iFrGhmZdblCl?Cq$S$#U9nU;@!twd` zBa951ihX+XMVv@i%l*Y2-ZV?aIWX4S#ukfgw{@ za%$RqL~W@X9(FzD_Jn4EN%lEX6-EV=!u^le=Q*W)7*XkkWXMGHd)d^j-N4BOA96Cg zUI8W6Ujb@X6seX1Q=)wDOINrgO5I|kJ0MwI1>2*Ya$B>!UX0tF-9*VWbKF?RYd_DZ zq^d|0dHUaiLvp378lf$Qq`ruA^IL+s?Z(>BVffqm%CJW?3xMbGUqTk-7~C| zbD!f;ltXkb3ZYHp)F)TBYa2zxTr;)!Bi{7e>x(cM#Il`Z#X+|9Z!jv3+U8SlcWf4- zzQv(zxRzXrc^Y0uj|tk;uj~+WR)y5$&ReQmPm3S*ebQq|i50iO`{so(B%UG!zP|hH zwu1-oA=FHX9yW!xHGf*aI95lWCr`M>=Lv_ts(ZuDet&&&q?1^(8+?~RH%u7Hsw@t> zOS6#0t6l@4o+P39zTpW(-Jy_m*Pp&zTQ5RmZ81@!6rO~0Vo+&iLY)Q=pY*9kCeH&+ z?sGvz)nY4HD&o*8#;XW&hUiEf_jjg3q2nn6|nh zko+LgsncjAlE2khwA+kRjggnndl~%Rb2V(t2dd155#zC=h(8=y*s`Ioz%*byM@_GQ|bSnu=?|201S)vX)M}#o?PvA5E z&;JF4ROmvxqaYLK2hI)$%k(b6cLL%(F0X#q@8QfxFbGjxudI$y>G#(celn3I$lGp= z4FU$Oe8p_D2<)!S0@f{OGI1gP`$2fvPg|mI79K)UDl>hQYR}sS%eH~YLwWx_NXmET zMey(_grMAh@(fr182Fq`tFvb}?S73RA+u11iPwhC7N#xtjpoNLZ`Qi*MDwTQAv{ax zq)teYbV-0hxI0E*6D_8gJdHeFyWVj4f+Z`L@B+^-ksY3IaF9gg;`^ctmoc&Ap%lu? zNYC)T^J03i>k%YA?V#F+3#sSFHciSiT>Ks!IjeyaDS1Ez^dNq8=4CIMOS6N)G<)>+ zCReEza(Y&Ii?`KS)fZ8I?n^NBj&otKGF0wxLD|ilh1t`Lyx@5R=V*!?E;%3Poj<~1 zN@`(9wlOy8H7{Z34=HbXE-eFEYGSf9SCkBJfynnUJHepgL=v!|o+ktE2 z>cgsg{v%96#(P1A;(Z1WN<&+Bi(Hukl{!LP|5|>PT1OJ_)8&z)i2ka{6x!18H?h-+x2#L=;uXRA)$g-k( z0eLRA<({U{V;_s#k)zt5L%2|tDTf>PzrL`MAu7Yo&xX)es4e%m&68c%td*^mt|Zbb z+@-yw6AO81FTp8J6}utXSId>UewF)cJ_L4AVR`3gFSq*RXxP&IAlV?Y+3ObC7N8*9 zChMGH!Og7Xd)zrAbEV8BtF@p}V!edQU3!Jw<2d)icQoL2iTkpu*|F|a8LRpt>gRkX z`{BEt2ZNMBafge^Zr&`s-X`N-lnX}*6^cxJpc+Hqc&(V8GNxkJibfv-FKf#G0Y9=U zl{fz=*_GxjRJghe(U_6@d- z$x?Qve6Qy!hu1U2LuGT=?h)&D?U0swkW?7_FQ13d_(zE=48}$>jVUQTg}W#z$dP9= zetf2p915es)Jtm|W#flIno)29(#aqlQ8`e^kVny&d~IUnoh^jlHk^+7z65k8%j{}n zrrKT%_SV@gKji7%p0xs`H|X_xfFezd(Wa;)38R8mH70 zEZGZ4Z10ccdEhcZwF|r?v_pEw8%m1`g;_J(PJZuH_VHBzk@ZFPd1--)jfd8jx4~!Q zxwTs~3n8b;Le3^^r^Kp41XGDN%(KdxI&F{j=y2@`WH?E;mG*zz61#`!suxDeC{3sH z0r0qnzFBo&Tdm{uUGYEP(3Xv&txj9uuTPJ)ZQ4p6x6lTB2nl^Xga>IRIb^3-hsW^F zeGQ`Z3)wXdq$%ODyj&5dv5VvfyG#B44+4PVk)%Qj;)OFGuv1okQEkn~@O<#DfCT9m zJ4Z+!^ish9FN262hr7W7QA7Po^AEqIV@WgBFI2PRL^L~EnV3UOWyOr$Pt;e=_)qM8L{bzId2kWqJWh8#&-W zIYskk;VsO#myX6U1!rB3r>bnj&B~ct1yG8&#rIOLhJ7xlJ8F%&u(_&v;E8lt#AZN>upWg(ubEdj?F*NI1%bE=v$sT) zgQg!`-cjCGm;XgHnQ&bJKaQiHKTVtcIii01fO@e~1T3c33T%D-!`?Lgk$W?$0w{md zM!&1Rz|q+$F=SPT*H*P*NA@b^^J>>L3kjDv+R|Y(3$LZ@>Hu~~c>Qlz|3+~7I-a1# zNpGu$2UmoZiJ}o??vegHOu7qoKRR24JCIb&F5byP`PXPZ z2##Qpbt8hyFMwpdL(s$bSV6POa}<7;cmY34LCa^@Tvt(sCTTTle*TmS<5FKd*(jEI z1+CGhLv08#RK4Tsu!l7Zj zQ~FwX9@J9-^-D4AWJqA37!MF$i1tdcg@@hk2z{9d`70H>`U}Sg`=;T$b1bX2hSb3k z?zZx|Z1d5mFZP(z5=*v=Q8q{z{&u`B>;cWfTt{>=G{tj2n?lLRhS^9WZ#jy>`)N8_ zZ~h7we4u`d_|VCJ$)!zKf^Xcz3nWU$L`>T|mmgY$o8~2Ij}$>C5GVZK4m_ zoJ47^3Gz*AvP(2B9`p9l>%?s$^k#&v95Opy*xEwp(n3wawKJ9e_EH;KYI0UJmK8^^ zzF>Vz^5{Qnk_?3g^PQ|hyKS?;ec0J48EnNp0Zb5kLqPp0HbXk zri3LH>`^*}SayPjz1EhE+#Z%#m-lM=jV>8*Fen!YFG?Vc#Dyc+)&UAzPB~vcEfahS zr$)MqP!=86$9AIRAz?Rrla5EJ`|2!~H;kZp+MgmTB)i^S93M9wtoj02X1v7QZau<= z76aiwVbNG8ng!Y0gkW?M5>&9`7_=0nF$HawJ^f$Q&<> zwOLx;g~!XLS8+usTO_zB?xr6p%r4mo^ zf+;eL(QW8W9cxsH^#}1^Y-9Dh>v#YnU3QRn3Y|rnq%)jWF9@3VEPV60)fbZdgc3yl znt3pI8C>6KjoE{mg=0D5In6I*`G1g9P9;JdxX-bQRkqi|IB&|)FmEdETdh`OhgDzG zTI2Vto7#fzjo#p;n-2kPo8i6q+(8x@r8#fly9w`|znbE1sf8uAkcvXro{a-R4a;68FhU`ceUTPGMBKruWA|N(r?1N%<=J z%~za1a!~UJH_BqLk>sklb(R~fPtu^1j9MpnPOHLhqtU4^+;o0R@V4{uHn14jbam~1 z+1P`ch06ATV14*;MrF?MJjLZ0eu7Ax_7UJ8wvMN1HVJQ}sPt>x7QH5Tc}?K2DPT<> z!YNFisMy*?jhSuwUtbux`6$6r=~&uQHB6MxlC^Kv(mu@4-^Swbo|Y;YV<}z6KaV-> zlohO9HYI2D8QO_DDqNKrtHA!jlTd~l5Vx;aoD@oS`W-mX zs2?o~K+!Ja%$NRa+iRF-Qj|vIxv*@q_j5_GQvEnja`iF3TQ84%oj&*347Sm7c00(T zf1|3MsxYYag_&C$B{2IHGhyg6wEk08pmn2JFM+QKJ2tIph(6F0wrn_jea9(HIAPI7 zF-CFF?v$7SM?W^5PAYakA~Kh%pb_#ZdbA^$=1N%`S7wd`86kb$hpv;XR0g#6OrImU zR}vkV38z56A#GJfoBc<(YdWB+oBzV0f7_~tZyuZaB3O!5mf&~C18o>FsJ`Qhv4=Ga zkRoo`BC6I8PcoMor-cfaNeh}`u-*}=iE-Ne!!ljLWhA(qIUk^V^|&V$1*`*#G*4#1 z-`LvoWEoP*pXi^)3w~g0kq|qsoJnd=+wkYs7jjHn3F4}cu&s9kSC#3uyEF?m9hXx| z4)3atxz8+htA0IXR-608(lsD)mdl0PoOtt!jEnc?7hRTOvs|Z~!HO4m>99w3!jv$f ze)Tz>rgw{!@EmcsAF(5$P%wwUcou(`?Xc`)PoqU}CKU?ye{bzqMp>odsV~HSo~ph3 zp*P?d{C2o>?B)$aWsLD*vo{X+vLRec)Fwr@+57mVini3tPN$3Y*d)U3o8r~ao01jL zg>N>WO7(51*chdYbc`!hOVVp*xJaP;V)cjP$AroRWCzXBXOU~NX{;6F=HmXG7eQ;i z6!)0#x}tCo?MC_;0U>mD zwbkyAQ(g+)^_X9K52SMAXsUKOo@6$lX7^DX3CxSwcA?;}j&SCEK_}4| z7K}5r`w^B`!s_?c>t*d~AnFU;w*}naHp+$u!{9C_-|pTltQ}Bq7<;)#!VwAM)CIpr zZJk7PIWW!#R4^Wsm@ZWx*AmW_jRP77d1L-T4M)h=WEaZDc8FnZxr~bTgfc@ygqe#x zd<3c_qKHEE9hwvcudHLMBGz?5#9b@a?f7Mp$r?qBMw*A8u1D5xWq7{9MYy$T4ua#Z?v7)i?^3Cjb3E{-6Kv z|5bf3a*IitP*<#gYlTCjKOyMFcwGsySMLJWfj@?XE=tPjn44_raEmym+xN?aq* zD{AfC(3d~P7_PUtd9s9EkrzyUX+B-Zf+r~yKSwi8x!rgubs8?Saaz?-R_uCx!KL$1 z;$Z#BINAy|l$6bowQkmWzDypFr$}emwb@Ko^74?TiSYi!ar!ijd(<+P1z(9M9Cn+M;kWFW` zEn|My%0Ifkn6POq!A0KYXTxht-T<pUv*>*R3eu0w5t^c*&a=j=kKdPd7jGodjvOCWkGP77$FOTv8C{wa=} z|3hRYS|`*Nb1?|Y8oO3cX=Mwu?I6?_Rx+(6h}wLJ4F-m--Oaw;yIBy!vG+YfO_@W` zcXmT4oO?gw40lH~4?B`Mz8+cbAT{H#72*7sb_Im2-cxFw@*D-qw*=26m6SQ^v`L1G zpQ;+1ie0ZS9#%)J3>d5JN=Me#tl{DBb7pOtwXTm?(TH_*)9~W=aG$-+G!2*a1`5)Z zx&2+g%9j^mF--5$9Oi{3@1FBpU=PpLy`&)3xOGYeI;z--1j0CjM|t|Y0^&GJ$g4rS zOC-(o8yFy?gkUzcCVTOaDyrh5Xb&Tl&Ttu=p?4O?4^?4fo!e$)>I))2*Clr4m(G&G z%OLoZ1!WgB3y+=XJ_`47fau!Tm%`S%A2&e|GXL_M`x!DI28ue zmb0z@s&i|1Y8EiQr0Mr4)_hO(4JheMbVNj)d_6kUA>asyzFi>E;Mpvle!%vi+0~iI ziKRtEXp&0(VSk?Kz@#9Q^(eZ^xV76al=QsNdApB3543$@e|YXy0au!{=(?@Aayg6~ zg%+t(Hc-LyPRyCd$A0rHN50W3Q~+J-U=3(e)sAgAaP`F z#?r~R7a>J^ZQOq`i_(}EM-KNtGRd3+#qIHoO31h-3OHVNjUB1$x}tNZ{NruMq`t6S z4qq91m#y9~$c78UX1-?LXF1sOnuS9VbcOCN0VQWrQ^YX^C|yrjUW+{U8nmXKtgPl# zmAJ5}QxNA;yg!qq9bCmzP=_a@3+0gjGn}Xbz&y8Ky5r0qyH63#_qww#tA@<(hsXt` zM63teMuY~JwnxB=EC4DOc%yk7i@!wpk;BxrI2Y#fc|x9^`+Pyho#u=|f?ht8tn#Wx zcJraEFJRq@DB0iEXub@Y_9ocL^0eDF3-C5W46i^K>Lq6+%yD4S*v(Gw88 zjb_;O3xzD8?ogRH!%$N*_maq!4ybUzg%ghB+xjbm@5&`tT0P7$nh z3FFAArx2K5oc%<)Ki^n-vOcMC3Q+&?AbLy)=vCu_Z1ywji=m&_KX}6++EO;0>}*c$ z*3Cl6BAN3d51-@&x0LpAhq*lYWW7~dwsys?^45InrZ_{z{fQwGXGyysx0W0bTqh+q zNe#}e_R1MCx(7i#5p5R_niNERGXaHDj(T!@(?2@OoM?NA0hG+uZ-e9>ljc8jWv=h# z!G~;mWjj`F@al`fw-FM5@#Glr4AVPXd3Nh&VOj)6M(}(){C{{gi`l(}R}10O_Y~Nq zuC8^<_%?uz3^Wllp|TJxt^IWTc;30WP$;`th|@Bf|Bc{Z*j~!xo!SYVSwW}lhVl)j z^TQCIuq~&Q&CsLV7&@AVZ3MOJjHo`r{HkWM%9zv_s5-MH8_stbdP9o=ar*^h_iPpd z_lIl}p-CR6lmc;}BjF$W@(LF(7g-pt`faZ++Id}5= zb1|Re!HnV|K1}{RuL_$=Kcv36@$*rFoQmUX>)5cdZYHf&vsU|cmP*j=I7*k~djfAA zt@$`7w`_m^vkTlz(6yTWJ{e5@@t-*rOu{XB%iNwQ!;u2S3$*X%)Dp%H{5c+q)GJm= zEl|PG@ZjG9*9pbL5dk=O_SY#Q4T;+cQpGq^9^@H+0vc8Eu<=i=F&3DvG9YX|sJ6Tf z1lvur-LF}2=|m$aiD7$qCmN3qGLiZMKVHtl{b@Rzwa}4#%^Rp=D(!T;vq4UK)E)_sOM0kS+8Kp-i(g$|4EaGU z!%@|7w#8orQ>BTwdo>G4`&gYps(daNDo*c2gh$-zu0*#KU|Vl}R@;j|UE8}%FCVx# zhLt6DNBnsrB!KWuG!5h&RIoit8FC#_yok%;D9eRfcO%4F@psl2HgBO}y>U!!X&X$o znOeJBvw(7wjMDHi9izNB?aF~fB0@x*SlVH`+a-3(E)6xSwHvi%TVhMDt>pHf$bQPj zqd0V+lM{R)PV4dV@#0TMua8Lg=TOr3)t(B_p+i!kP&y=Jjgi}_oj{z##|};ZI=w^x z!+pCq1fm>esO`<6Ca3dX6Ac>=q5)?gWUXAwD-^XbpHe^LnEF!F%8vEQgH~VM<{fU-C&67p-E61GNbGFro%tVY_|9Fhw z1;tnQ(!hz|8~Uol5BlehlvMak4n#~v2K|Pgjw}C!th%S9a@1|B@vJYb-=;_R&zmko zr@aYww=V78%|bq5X|}RsL4F9MnC+;F+vy2qq#hpKMZaRGZ(5(*hb#FAzCHWX295KD)06ad$9R?Dq4ZRO#u^-G}4sS3oC8d1S;( zNc3NePTvf4X+PDtDMuS?CrL#5g}6ARfqszm+}{F^Rma;F?q{ERK~<5t*$~wif-=b^ zF&h57lVLWX7&d>!im|6Q3!cZU?NA-0=IVKj!*)*-`w-0>;^Lh@rPGJ?mNIt_aWY%E z#$3BDNF!1+fNxoYR@}U5B>8HZ++|cYIP7T!g#s?U-&VS)&^r_4$sW~hLMcJX=guGz zTDEK9+`eCh_D~g!TMS5jk?AJ51e;q8wL!xGx$9+NPiPkSNcfozvhYYScoiknr-ZCp zC;^L>8c74pdhF#_2#LU5@d9E#f%aL54C=9M`N{?#$DH_3g25;jU98d*33Q|_3c)Dm z_I;Vx2d^=~QIMB-*7_nh6-#?gZqk1IlvTxB^)aX~+;l=q(6+@e8xjm_yIBTy-)14~ zK+3D*jdc4*cn1)72H|58y#gt>&2-JQK6+b^bgh1dutP7nc}iOdg~rh+*uT7VyVHWn zat7$YZ591XRqPa8Rp*cVlo5XbW9%G#(8294C1u_1FBM>*t16&2_{sG}kWN|&f_6C0 zw(^&Vs;~;|F3rM^B;gDaL^#i~5uq^Lk}ij?O93qW{ew|7eGtyhMMPLM-fY;=% zb=*b_Vq7e$ydG$-vH1QWxUERvu588?U~1R6^=yKlvWP8xtfa9z-itR)*)kJ721q9a z!&|rc{?EBU&b&aGyJA`6peooZj6r=7<|ebmn(chp4H-+kRa+f)qh>+u&v`^*Yj+@V zwvQv~D+<{g#c7Et=aRT=7Kit(j+c<;sI=a`+2vkRH@_UaR=iyD0yVqx8V?2@kgn$1 zu9u1Tefuq-qt1xI!I@Ko1jcw0^?i2WCo#!<^F5tbc6aP(Bt(YArL)xbKb@6$L9(A&HheNCe4B3tbkU@rTyhvrK# z^d-s*u$O+2+bDffTwjp~!WVLytqUQw2oc z6l6%m-sewt@zoR(r$}ogXX}BY<$|I=$9Zn1Y*DE-Q9z+~2nvgKPuuTHkuPwVX53^l zsH}8bNt+B(| zXwdV#F;vf(a!FmnLXInICsiDS`eMpPuWZ9;}AGaDAGE$}qCKy0c&>4bGn86t>C7%8r)&xmA91eX%7ISYoYqGU5gTL)reD z=ijbv7JzQ32X-O=5=j#0wzN`Sq9fxn?XX{^SB`zJHH;a>$CsI1!4waG|)9WQs@z-+Mg_**Bt0oem$;$vuwDc&h0K zRb3<7k4k-k_7m-y-A{rc$52;!x!A3m1wtt~FdHP{A4$j&rY~nJ0(IVWL84siBq>m_5iY!?1bvpAGhy;Sc!pq%qG z!PSnH0)Q&HxgLh5@L0UWsdSQUI$6oynR#`nRf02oGHR6EsWmKmIvmdEn*tJ^5+M#i z*&e6qq^e=7`b+BzE}gd$c$JN!EoVc>PG{3@-z>Bok@zx^bnPB;%fn$5ej10L?(ZcV zCOkwPA=9kfY(y6aOkQ;fw`*dy9g@x*IC6ddR2L~5Pq zPItdaJWGdj$rSlKj1XpVhq8loMu>hsdq@_vAI3joR6(m19;rGG^~I76ZkfHZM!TB< z1B&6W$~v*TGz)IjT^vVFJ4s~8TJyB^Il*QS2RneEmWecO#xjU)H_V0uL(?wSf!(!P=whg)v>P|@T;t2HX(P}g&=x_0>=3jP_u+so zBDy~F#s6l%=NMUHy*lw>;-uSi_DXvVJSc`q(Udx2O0vFJSa9qVOs~xbqJA1hP^oq0EX8U(4Rf zTjE{uO@j!9PSm<$b7&HaQz0{SA#hmE9<*MA9U$bdBH`f5=oL?Pu0QNQ>73$H3gD4( zK;5BX>b^OTp5wc5LQJ9a7|(~%xN=x~^AV^oO#MWbK(*4)+S)gK+EUrs0ce>PoZwH4D8Mw-IU$k5TxwHH}2lXf&Z(xI_m) z>9xyVdHyxxVnzDJ+5CuEyM8V6b}g8n8>7 zg&7eFEX+ewvQf`(>0CnTytfwLqE%orjoOUqVJ9F&w|n+(!%eVjlkx zhfhEXUja)Znc#Y$@c6<<$r0~Ab2gQb_}?@ZZup}Zx1@Cz!%3-%p^blaeWAn5l>lek z0k&0dfZ1m5?QRXjQ1+ce+J2OTU*hOtxM(d?PNgWz2`hO$j{eqM>$1qgf5Cxt0K34o z+0BMjF}gn9b0y1*=Q6qgo}@R<+K5tRlCiJUkNQ{HFy8xWuD6=VLsf2)5voI}%8#!v zoMaYD@U+bc+v+zg?O^`xrp*72qfouq!S7p$%s2jW(2VH;qO&VRfUnLFS=x0%i#46<8A;k)NQdy>}JhEn==yZ{k-3U5S=x(Ytuf%k{eOzso8)UYdUb|njP=jYYl=BEB!Q;tr;Y<{{+~QCm)zUj5*22YI;S>+> zJQS|*+CPeyl^T7oDE+$w5ncKJxqG|jwvlCB^L~B>ug;!{SnLyLcEpYepEG+0C0UZK z535DlZh0;c5DAH}NP-O#lBr)mPi7U8gbcwVNV#!vu~wT@6+Wu!WmaXr`KDQ_3;%C9 zOo=v0UBgrJ^_mM0ovR9f{CX&DAsaBZnn=4`yKsTWnH-V??{Ma@ka7MwLx?x7we!V7~XIFS1 z8kLDt? zCsd9}oBKeNTw+{bQaFEtSC39}F~s~e#%@CAo?ftNCS8P1>%)= z_Jhz|EdHAPw7~!y2n=jHoO!!=yAY<#1E(*n0s(tTf-DVIkPNPJ zG)XgPd#;(GUe;E!|mK<>gqcnO_VYgMx6H4y< zh2xGWXRR8-+nKwQjzfYaFRRILB-?XROnT$^PD;H~N=5?OSk8A*aF{_j?-l~3E&!Cn@CA!j z($8m$Yaw^UDSnK2lIT*jZuuj`ezaHDvzd^MA;r#eN2E{UipaJWArvYW?axg z@*E`J{Ht>=-a%*c;Pu^b2}$=Gl7l=*>TU_QAEV|1GoSei6Zm!$Vwf@zZn==`g6#sS zsKvwx4U$etxukd)1|NlY9wa1#OngSw%SSG4+gWd`U>p#x(jYID%N}rqB%OIQl@pjk zE{z7t=gYK4QU~WeEC{KL`xvH?=#YA@UlF_h@tnSmKCxcxDoJB-*t=)ndG8?o>7I;D z*`$(e5-#Mev?VE>kl|0nm8qZm*VW@U+G`D^d36p6h!U4>O#cW8j?L^GktrFqnMi&C#11HA?s(dnq|$ zFsNdx6_4E6^*D(mEc*d^$MpumlaNcytPRfpJGOSQNJO>eYXZz;V)4Vx4)bVQKO=h>W%@^cLN zM-jWJ&!)YO;xT%y2%q*iyGD=t-Elld5q&9oK=0fQ;ex98Sv@LT3 zQT2RV`*y)XfAA?z_$7)LkgnD$y{|krQi8f1B83jQ*VHplNH^qk$=UD&c_T1cC^>PI zB2!;H@X39>*`eg1Zs{VV0PbQ;Oh3XoQ+Eqj|HYaM4xOY5BlF9_v_))-zSa!dW!i;> zBl$sIW&x7(Y=)fjR4*73E}KJNXKj{)pr;Wvn&|TFQ5`p4M0zOFO&bDllEH`+kt^Ne z0&-#JD;43zS>h;*1#{vaKqPI5Rr{yXi&I%$ti%JMxjUaRl4m|paQb8Jd+Ae^ZHlZ0 zs@&lMu~O~Z&v0zAe-pnb?wY!|sxe5-1(b|%RgC+0OoV~SFu9WjWtVLiMqyVZT}kE$ zzGhyXRH&qRvdO~b-v@GgY2&MK)y(V%j0K8_%nlF}54 zN}4jBJ20_G%7dUSan@d0~7*JKPCB^*W( zfxSL*kW1(9MO-)4E* zecJUKa2*c;EzIRC=-uANQ-F*2Q79`)ig|f6>l=&z+&5OZ>Cuh7Dkn1C)qWs;7s{FL z#l!(_EqP?}72+%=w$Y#r|sR5|(t)JFPM@193T-g}bPc2RCU z5Y0tUKD8APsx#7t2m{d$*MVKUU2x+B72mRKgBgpM4{6Yk#<$&lzPgcGWb6_@%#%r^ ztm$&FJ@3%54joA>xQy@OHFif=s36_5to{nHYqbkii0GndkF+*RIvfU{#k^4T z_B)RMfPhz$PTWDOG_tI%JFc~a$MV*Cvg*&0g(^;6p?z4eP3~SYoqBHt&(|=MOmIBQG`@(f&Wj6g-`}xs^#aN2Hzi#*4h6<0dkh4 zcrvggR5@RGjxYH*P=Te)^MaJ2xHhl9?hrH=Fn&HO08@924F!gv-7NyUcDoRUAQMSo z;>G)bg8&XiLd0uTEsTAf6YqFo)22OGK6DM=f#|5p-T>NGwYP?qw(riK+EZxruPU!az7xpt5-%j393(lU1m=kgRj;@~?8pRy&{ zF4{`v2z1Xl>KzQgeZ*^Il=@c0UCMBSY&Bmycb%T^N;8dWg+SoT#Y8;|l>Q z*RblqHC-ud=fwb5ee~FcMLda?Gp$gz<~MwZLy!q29-`cw%_=-b>SnGfjJVj9dq4#o zzQS2XT6ZrsIc-rG*)bJ#@f;U%>GjDPT&brHJF5H%N`I^z}a-1X=iehi!i z4If`evi{Ps>$eS`q2yKvhbWsmzGL?Tl})4sVD1!=N`QKLD>tFfdgaa$>!m<}c_KB7 z3i#uePM|I)iu4l_TL2h)pj*^R>f>pfS86VzWQr;P^{e5u1#AG>Wag}Q`Mph>GYcGmmB9+Cvcwu!mme+oSd zO9t63YQ1rDf#~M1!r*N_#J1)QUbQFNF4ZphTxXnQgt~qNy?z#)OMEMLdX!s0ZAdVA ztziKOs!1?B0Fq`ukN>=B#0zJJZD*EfAWY_{-h4VTP-HEP%D+)K&W2rkBy z6NiPj_FL>K*>xrE8oZy}L_?*}mU{fwdbwY4?FAdi_p-26VHNv(LVT4+==5BK>IAhn zp%Pt$G5a_1#~+77iRy;4L>vicu9>>v#F3W~@|9K>aT-ptg=p53rxZ&uF8zLY6_mFf zspi5apZyBM@J~&aq0XTB>cwVV+XYzRjDy%eC4*m1KVY^?oC;Dqj{6t0S@fK=*3ywH zWFyzlW>NYOvmslZnUXsw@&v#Nc$GVXIK90>%Z0Q^*Vj+}fP|Rv+(7A$RC`Vc;5s7~ zj;-oZ#=Nc{V)K`8F6wj!Dbr{~n=sUx`BZtbIbf?h+ADl<0 ze52+`OWIK4i^bGt_GcwwEPoSG!H}COcFwsIs!*I-&dIMP%&p&kO~$w4Do=$L;uF8l z9ELt-*FeUjo>hSkYa@*O9M#&gGsCT67$J9TzA`txTD136eLPvg@OVs!O~5#NjGzcq3CwYZVb?16y^y0YC&+>-IX4`)R3>b!JvKCtpEnOle!DItw)47 zt@WRMWzBw!Z~czWeZG#n9JaIpQbW=kN4I`FPUf=+s$;})E!P~Z52G!IqPY;$$*r(P zZ$Bjl9)ssE7mamk7v8SP-@p4FM!_}wCNKNjkN(a?kf|JAJ#U2aR(b6(_?ppJPf<=} z;m<5$mxj}!H=RGivGueh)SlBgdI(9{SffcuDdhZ76iO$t$Zvzcd6e9*zB$UKqo_7j zFu%j=UbeXaVy-ISv)NeMx;7~6YJ%;`?LrI^5)Uziz;MJ#h}w~e1v zws^DVVnrva0xg>kuB~dr%j;&;`n3xUBHN%c68+_r6U!nd_)pE^l@ATlPpAserF*qP zugZl`x^FGyEA~>@@7BA*_I57hrN|EB+Z8+kRNAFlb1Wn{+V9vYq&sXX> z5Fte|t-cR#i!o>}ZeArCQE!|L3I?HFECai0y8wn5GGbxFD4<*r+3tEK+8<|7O=U~< zEC04NRKrprE3P{cyr<5CuE;Xi`x0jI&!M>5OTNRMdz{~6hA!e{;R>RBOmM;huO4Aq z0#Ne5gCu*x!>x+W@%P%fhxyGn8HDCS%gtnkP1@y<8!QZLdem7hp^a9CU9DXp6WK6q z7Mbp%?g_Z25ykb?L$|;G#cj4{Az`6EW~14^{pW|dgj9Vqb=2)sa8f*Vy#aNZ!@=^u zI^8ZrJ*Pbg|7|WxxBW$i`Z^~faqX>LfPVku|3L~mf+tur5J z5}vr`pcr1aT_twKc0o_dCvyS>WS`R;{`oCYR)*8VC_NEsyN+w;i~k(Q zcL<_&KT-8kD*vSS$phXHDtS$abVcLgTx_b7chX!eHKOSfj6ADkD-R-`d2c5vx&i$E z)%jC|y`hZY301&dkt;IloZn?jF_Q<`K)=OQo+S8TOh#M$Y^UL@d*}&Q$QAN{Lg$En z@N_(@{1#kCTvjpPJ&6(Qs3*R$**G>ArMeUte(es~Q1Q2Un5z$aV=s($GgT!)h= z7(NSs#1s9hRQsNtv*?RD@2W({Ph-^bT?~+nr>1rL0%< zt&90>Mzp!V0I%^YHGfVG#P%!ky7g=Q+68_oxpOO4{umqI$1g=OGqu9O*APclIpSSA zf8^D6IA+mx!)uf$gLX32Wjx$Jcj`Zv(QO_w=`BdM*7)LQsF}snF;C%A+2onrFM&7a zq*-I~^(%sbLYEE`$kbw}MKT!B>nXV|WH%m%=HgUlvcgbrJ;a6tL(opwgI&H|KvQGG zDK;9$94-Nz(7ysWWu<%RRXax$Bj$mY4?z?Wb9BV8&dqXlu)PZB_uXNH8)!(I&I)4C zsYTrXIRV<6zH1}M?1vc7vv`_I7lo?Z93S#KJTp+3*Y;Uu1;jOvPEo0llLEl1nW+)OIQ5Tu!dH8qB= zfLSif^|Cbn>ed`4d6rK954BL_U=jZ$@^eQ>3t%FcwteOkT)xcLFbJgniO^+eX;JnRhv#YfW zw}0`IsX7aC#<)tTBtOjKIr%xX)J*!OJ|vT$K22Ft+`Zo2q&l}!yhh6P{(KhAA3)J6 zhZxR8&C3nodIyvuWzUy6tydf>bnoQd3&?Td z4Y{q(C+?d?G{!2qBJ29yR}_sGA$8yJ(va>Gqx&i+wthIh$E;2HpQpa{=8x=g_nl1? zO58l2jN~xb#|~4kW*&BqYNs{S9GT`~O=r2{jQ87T#vo$X)=Tv{7Ven|B(Aglp~d8l1Fb`Ewz9uXlgPj$z<(H+}2vBbcurm*&Dx2DECQ z`6dHyI5CLrdePV&+J(eDlFP$5AhbVE4mDJ%%=UqvoPWCZ)oLy=nw)$r>o?&YoSV zU5o#K!qQ}ZQ%%feU zU2yn@*y1QS4WkjVY)Z-g%K74+ywAW9Ko`-=_27l)nKwtwQ3<09w{M~d{8*$72A%zJ zB6f!I?~+x=KMIIA4duqT*5x@A5Bq6^S`nwFHH=&-@dUkn6p8DxPHMkrbK&8JuF7I; zI=HsR4Jq|z*sjwqte`)SFn1R0#ob|?I!YvjkUT5sVoADOJS;aIdz6Q@E|wDHyW|>L zd1M&kby|#CStUO}lmSkcTsI1b$>S=-TmTji3i-nmL8wyA0q0ws%bOnaHwo}Q`oACY zd4Cjlzvf#qsY?4-#um-{`?TJL(Z!DSau?zi#3^4}@Tcy0HW!kzIx4KyZ=4f@ph5X} ztU$YKyMQm+2={ExPcxUabQ!sFwj?p-Rb&=Wf zh4{%4uD&MVm^38FUXCh#4E!Lb5E2uiFDdq3cjIagQ)5J$i#(m(3hce%L>O`mcbhL7 zyGFavcfp2E1t?FKS#Sn}Dv$(qEz`;2u=fHrOB!k>yKl%z5OUUZ*j*%D{v$>&o=W8H zSvcD5Vb^Pby8Ul2y@?lTNY(>xRtGy@55 zbpD6x&GP0rL)BQz1+jO3jQCrgfj!Z!9) z;~02dX7O?wI=FoVUOqNR(m`*nr{K!vh18i187w7T&l;f zsrKpfb^P-|hP9H0#~87%l85J^C%zZwhH(;_2jOFFcgog7(_Gl)GhG3_t%lsdV&Hp= zC1Q7J7bH)XB9)mC_lwY{O_N!8N9-CSc-p@@3;z9}+{Udn>`23R6864G=4d+;=$p9c z_f06p27sg^b(1y_k?W3yB@O3UJyC9sTCIN+ca)QFK1?5z{OqPp;6~n!_^PgT%Wv`e zw{0%0WWp*Swds)B`d&kkt%kajY!_`8XfCs4K*TFJgvQvL&yzvyFr!Pn>rzDf1idje z<@j0a#3Bb5t`po}?ZkKUh0sr2aL*OkUe^))b+!$UCWkS7P6d=oh5Q*4DDXQ;cDgOF5QVjWr}G_#qJ2=6E*hT>jy%LJQ7&UAol8;8ugzTsX>FsDR_!&xc{n0Q@Ga z%x=>zNb8}AMDO$wML?Guhu|j~NP~oQGBJo%1wz!qjnQUVwI{})bwXTW2%jW3OYYe` z!c;=tW_1cDrbkr43%X}0$X1QJVn9Un@#QRCekg`0;|001D-%Z)l?gOUC zrvM$sW6bwFr|lHr>oeod+_NZrUGfQO1K#3F0eb+LO-cXpU#V&#u0L%r0R{Zn(l;@>_khB3qM*6YkJ*)G^#3#$&v zJ13gaeIQJaO^#C%6mXSVpuQZu(h!f&J8IX7R7+Qb{PN|qFu#{y!87~dxdbE$J@Xhn z!?B2}T(1y#8t1~>hF?XaB*!W~VF>9LCrVaDRivBiFhN`e&~($ki;~1~SUXL<_HZ>B zz150`vOL~&E(}=))?KbEyL!8D%L<34Bbi^QE(CpCO2gnKK?z!Pw1p%NUBo{~i+zSR ziP_5)6(XERh3_Uj9Q}NmWRww?Y?tRo zuR>l_+dnsR+8$U?^BKfFlAw|A8Z`*aM)qQqED|xSL^(@*vOI~@W6=jwh*8jGI$%kB zjIe0^2nX*XFHgCbEG72%emDV!bs<%M5Sj}zH=z|LNo`PXoDB;Gm>n$yyKuWeMsQm- zyMimRS0a$Rh@U9IM@|IgQkD|tpOVYLH>HXy)`1|Sc96KP>(Ke!W$WqOwW}~Andid^ z%D#}g32^$@f5F(vgNT@zdoBf;JM#VIHjqbp{=?&%YmW(qjU?H-ym+5K%rNhXgl+T+ zXVD`3A<{_)92C)Ns_om!pR2}rH5ZUN_Z4{jJyT)`GwAMSb=h^>h1?!RAI9;099+Tw z&b7;kPC^bHXxe%yU|rym&}80Bwwz&S7*Tdf1z1`|Z1#rDMTeWC3Us`39Bm;RI(9LmcGY&l11G{Xm5M3E1t_HK z&R2KHEK4?#{N$JK`lP8It$b5w9oYF{R5#C;?yBfwm&P~j=FlVZ zd$kk!<}4yBLR-l8u0P72d?jY*Q2a~k`qecmHQu_pxRI%=K+R4E+)!ZH+Gr(M_jZBk zV6vnNP;eH{N9?E{bI4DiE~8F}Brjz}j*#*ah1Kr#`;4K#LbU zhxzG;ix*YAoQs?mk`8q=MtY(13A$HDEWPU3`A{;#c1Sa*{a;n_$g>Ly2Fb}Q$TPV8 z90n+w;2cP&$=IRB?dil}X9^9c9K)nq^PY%~A@o{zulL$lZ!Vm4-YW34li{_6Z%Eq7 zIs{l>b%fQ=0zWbG|oW zdiax80a(gs_s^ox(l7c_hc3#GSBadi_~JV$0@~&R=Q}jWWwr>N7TL*T$QS1c>PK}! zR%bk#3pkn83cT%n)D0wty?iBbj>u9G23R~M>HQ|_#ID&c;7N%Ya`L~fRR1qhgeeEh zE=5n~QTia8#VS^EE&=~@x15~31~xrw^@dVAL$oi)POE`is5>x5xetyZm7WWr6nBMt zOlR$UG>iW%a|efGJPMT{=S*o@Go+d=oElpVO>@D>R99ee+rc)N81&v`f!J-@g+|p0 z0glI14ansdTp{#1C#Pv3bpCQfm^a5yIGKT5Y3|tnrxKsV7=E6P!<%T#zkVw2DmkbA z)(SD%Jqlx(jbdE`R_{%kiyCI60yPx_Xe-qKvFTLV7237dPf3)Q6?Uf@cVGHJ&xzfY z#*if8`b%G|L_sHnilvhN>EcB$uWhg-uHp@mL}n|;-`977!IC4E83H1{@+>` z^(uDzI48U>iEHC$h9=v>ti1ZYn+pavZ54Rf%?R7VH|T7;4D5>Sf)W|aWRKh>aEqt_ zgOc*Wmk*yrKdYDFoa1kUa-*T`R6ffZ=1x34ExCPP_XD9>$xi8fqHcJ!NQ>JucTdhl z{vp(Y^x-gYP7;rVe!dJx3t;x1Dl+0SBp@{n$0Uk<%tp0|>G=&edF$q4%1>PdsJ0$x zTm6Qox(mRr)Giz$?WTtcT;*}Fc1#J5BDe6nJ*W;*a4fa4Ux3uw;ilNDqrGyU#L7k$ z&N|#!sCPU@7cZ1z#hq=gv#y5g^($@Q1X#&G)344F%Fuu(kw-ZDqjOl>y|?8UG#6kp zrxobh-N+j_RxsOmQP?%w1+b5Cl%n(qf#?&-h|`GNXLZOSzPLK@lx_fBd&s7F`&A4I zo}Jt0QK+8MFc=*YHx|no2CIjaSwcK7V2s<#*@h=s!>w*KS2ML!GA!_M&$u6*x#XUDn874g)_i9AeXc1;M@Uc z$&@N+<1i(IiMkhVKeiVwHKSrp7C;sG)gI=$Z|vs|A}JE;ZIG~ttb8@MRM}G>lJbmk ztX7_6?1CzCm~xWfdJ+8SyKysYI!et&8gEhowBI``1}%eX&6Q;rY8OVYviT@XgLCAp z((pT5LL}nS_y&cFCK)tKxFF7N4Ak?M>*@(yCQ{;1dB}xSzF}`poIA0yb$pgL% z&u5B<%2mP<@!ehGFiWR2&hs?9^W!9Z3hyO*<2`*zF-e?I&sj2F#B=eQ@DL$#s?O_p z27y@8ACoB}LH=}i5g`U58OB%1%~AnPxNRUxk@46)=~oK*iqaNhlUtN@t2G(TMYda4 z73Q?POTd0E?3btxUhFFE`qAl)qaRvXORB~pB^d=*@_+s1bO6C|s_B}gzd3pPjkxKnJ;?+Zu**c*{V^}+- zvex@G7skJ;a#d$&ZDkuiHkw=O-Y#I^rVnp}-WbovFvd?aK{76IK2H{VIGeMvEd_>a z+R>(h3vIeaxqlpmi_ZT!O&0%`kRb8c+(nB{0U-XSA_d7aC_eUtj1L^4*o!6r>0r*v zz8{ZpbH>wWb#g9~QF=qHPojA#g}(I~SJ`rbj20DqFrG1q)OB~hwT7m-n9Aq4!Z`k} z88Ivw6l<*}yG*+v>53MjSCrERv?djqCIcy*QI5vEsUje6+BwUxbxyDfvh}Pw*0&WNDHl^OtTd=kTg)>RRiS7U| zF);pxn3aO>GP;Qde{&p;raoWiWE#aY@EGEM=J!D&`6@04K8Lx4I>!vk(q>OjoR;0G zwk~w*jnMz@0QL%iZ#*XkD#L8eRb>}y7dp@50kXxxH@rSnqLZU@ zD!0DRT8ll^)+8lFr04jsi#sVghD|xK#@(Dm0(};553$Q`JN%?y^cNr&qg-oouiArburbX2Lub-Z zQVUm=DstAkqpc+|PBrHQfGExkN3wOy%Y0~64**mMeMXNn@!AyN=pF^1syvvpi^MFK zl*gLiFY&8hgye+hpai6-0wuUw^_!DW^8dOnNU=W>-T+b1x`(-%L(^Q4`dO}k>Fx*L zU}N~**|M=qw+oRcBIqw6cIXA8Ni@Ds9)jcKF}O>TVGzF%%%DL>8g$}$XR_=Z_0IpR zfU5jT5~j;9uaM~vMw56bwZN%Chb7l9du)o86L7*lHj3`+B4~B5SH4Pfv6ByA)fO{v zQ*V%MIU7<+Ce<3XYt4TN(`*`0Z)rwAIq1t)Zzc-(L7%UbVUvrBl%K>_J;>>CJ@SP4 z;?c|B0$FLfqeFlYI-j#dmGqg{=++J5GaBgj8b8~-b8~T$&tZj;d4oAHSQvJyR)n=~ z7l;nYB$z}aFf$An{QsDLV{ADAA;uVdo<}d1p0HLOTFLf}a5!0bB;nx?Uxe-iMxc7y z=|gjaMG`^Gl)v_)J2|5139D#t+*>FR5NA;Q{mRcSB7| z7guf}Tu@5XeAdg!sjLSLj>EYS%XFK%k6`P(3-m~(y+rsSjvfRb>$GNh6v;$aSRrkq zrM$jN5JSIls{}a^E?O5@TfA{|;pOJA0$-ji_o|V#rEUn?ewyu??SjumoGj7+Us{~a zpMtY&9%U$M{TNZRei3{?(;KPdc%Wyn@@=8(((zq7#V*tk6fPA>K8hbvM%J_Q3!yfJ zNJ;Lke?0v2^0eCON`%F)vw7yI(SCIt!@63@{8KI(;|2uJqESlvfD3n%0vx0HtS86V zedT*!X0}mHE1;z~%3o8$+O>`i<6}-Kbqq|CmX{3Z|A=%lFe^cae1Yk5d`z z>!uSP;|pcjR_&dMT(x&wQOBD;NW(`lmTD*;!|{Fb@Ta%NrJ@g_F#asK6d&nj_;*%6 zQLpx^H5Ynrnksic zTj$EId275I^G8*v=#ZYemm~{GCYWg{P>j-W#Y8D=9?$GK)ITZDrXV>;@S;c_kJZ=} zsjoz`Vxg%_oYf2$$fb{CxR2`2th>3~YD}7orq$3^#n@J@wx0mQjG=G))neCd7Y?a- ziA=H7-%Xq`VpUV-jJSer+spT!o~`2=uRx`HO;@oFH>C*md71Oi)Sbe7EhK6|Vvl$} z#E-e6gB9Q?&!c5Ra}8w`!{lKp_@;gL<#;F3j(J_sGthS$KWW|Ux&W*-7|jKqPHKe} z+is8z5C*cHFA2Lv%PE|lD4RbZ)dByAn9 zpt9X0+6CJMjNVY<#=#zvU<*n@1!K`24-f*mi^{QZm2|XCN85Df!Y}EfM@9*Fu`>T6 zNJe(2u%;x{+ch1*sN%yg@tp`;rxMf_t8^Qu^CD|epGEe1@?;7T#}YLBB&2 z9qV?KMtAUn!~4-ROr_wV=qvc3v}`E#_R+9=rm3T79N}%O3&gF4s=2__;jTJ|z4d$; zvJAjGSyOh|b^%t*%p-cIALDhep`NdToUB!D8rS?y?l_h=FL1kr6UG5ev=;QH{%Gi_>s-7x=`v{#(G21 zWF%I@a3_OpurRdkWL?;0+l96h&OI6BZZhWzxr%zVfXc0PG;<)Q2d? z3%yacik{}I?h$2@bovu~lM>!8fb0}eF83IE!RthnZ8@C{o!euN4YA;fMCsa=ZIicb zE_|4zitXhF!)R;SAhMfjwd=MGBX}(?*p5Atvg!AXP?yF*HIYB>;dhMEBgcdbU#T;2 ztvp)+6)wwgj|g-3+&kS*VycW@^Ua8Ijfxfh<{%`wy0)aOW;^DO`ns2EE<#?Hy7~G+ zwFPZJ$w%tdlWLb}8z$lat?bj@#eVZdkV`hln`Fas{imM&mwtm120-uaN+ZHNw1FlS6Q3a52OLWF!T>C1bdKnA(2W2$%JJO(_$4* z{u-t;xD_Z$QK4q=f-BKn7p}GFPibr2vF#D0_VKDqg3{zAi@UG@CJ~>?@2-f_h?Nj= z-9`?9OJS zXTadH!mY1~$t=q1dd+&Vi?j=Kn2uzr73c{g$G3|vwkgmvPViwmgYzz{zT660`*$68 zC|Y&U^1Qke_r|T5iv$8<1WZ~PF*ojSheHjO&SlWJ2-vtQLl2S zJ?qzT{n#>EKecpT1TJsAVpA22-I1^}E>@-sradRIxv1+H0MYNrjhURrQJ}CtU9UEo z#|gjMC`1zf?(F|@{D-!2QE0JV+gQJ?hOD``WbCVO`CF&Pz-GYB7mi<^8!t1vV!Pl= z@i)>c0|qn84+g}b4 zhXm8qnG?OO)!&7PMP4YR&1LR-&mv!yu79-SLtjbJR>{xd>?gku#;_erX-J@OL#L>U4xlpK&UKnvWOkaE-bN4ZUP+tXEJQ8u74rG+Wv@$(VU-GWn{NG%9@PN=#g1pF`K zkm^0Py_U6}zqz3CGf{z?Uk#+K(-js-NbIal>(4V~ecH9QKg{Rh1iLf*9>A;MKXLGN zL`9e&6_rE&& zJRQH!yOhc@lc#VfVj|}emFpB17x8oy!Ksoh=8#D8C$}s75J^=~G8TJU{15mLG38k7!jf9OXE16)KDo^gVU4rtF@ zA;%D>z*t@EZ1J|u#gWcl1&($$>IMOW&1(ApTR2PR|8so}xNmGBSm$=(=QKtM|7$Y7 z4f2x|;tG;pB@_pDVB;AjN|vN`!L#|RHWxE)$f_I+8xE>1Z^Opcb8MGv7h?LzrJ-vY zQGVCC4ED%OM8A3!bS`CmsKdZFHL`rcy67cAfm~mg{2}f&Jqm@%xHb=?#^W}3*}GA% zuocx~Xe-hddFiZJvvw_96wM@xBxOup^5B@Fns;|3R*;Zi2qWS9DqX{Qx6Z2DdK77m z*H>8BddP5h*L6*qF-LHC8o8jYrg(6YU(4rkZ8R&_w?RLK94ko!k#oqE1a9#b)Fw!5 zHyX`_h)iz}I@qZHOf#%dY>r&!$5BSf6@ox$eon)Oy1T2| zFWOuvF>Mt{+J2~Q$s1f=HR0B+T}Zi(IT`Cw9#O&B491Tz_PTo^<$;2u@Ckpl#a7DA z)AL>(>9rP-QC`t|v;#9egXD4Q-h4Tkh^*j%@C@c~K7W2dkrOBuOP^4!kFXkq9;+Wb zfA+5o#%^tMwB6e`7ehL9RWRCMY;EBiUUoU-cJ+3_>V$}oTA<#+mk<5-L4P^9p)A&X z@^WsIYxxeHTHjQc`THMOFv;CGL9}Tc4ZXPF0Yr7l>~$3N1>6RSo)!hjS106-3ttW6s4nqdab%zA+_A#xGRX-X89GI-^*PAv9ffL0iUaM+5I%lGoT8) zH@9JP&*)8y8Xm$jH@=*?m!A`Y92|9RP&8lrhlCqLstb+>lA8?@#^P*+67 zZtsn!#^7ck-oe_ki?$2H*YS4>3y~1e3-%FSLG%I1iAm6}Uv7>Nb&Do)&0dX=QtD)Dls0tzar1_Ji5NDkY<8p6j||GLl**+OpxyJoVbs(e_^(e z3Mci$KpT%hb8+|UypG)sx8cD+R)1C4wc3R(l>aDF5cHqIneF)`q5J&Tu z_l|`=NW-l9=-Tg1ZpgL3^?~m}%DpZ)cHLx)NWyH>`E5jSXDv%nAA@b8K9u%9=HT_G zD3rOQ7oyT|UE?Sfug4Rv7GHDWksNvX5>g&{_M!80?8KeE)c`dYhHk{>GKQHLRmX&@WcQm4IdDjUvs9c!y`+lwd_(F6#50{qysAM=9kWCNrd#;2B*;crKN+F0otou>0YR9qBk;EFG$c4{{B>@+moA zV*GmJJN)&F(WHQ9r{t+`q5Zys<~>rka?5zC30jf7kN(}6Pv}YdNqc|IML?ZP=H!u| zg640d{CQ|e?0AEq-MFSYXB=bQsgT1IOsB*jR#eLVCuN_`r*ip-kBte>+ zt^@EmOjfaY0iQ!-A0tbkBU^cvfNySrec0n_i=aonmv1g0bs8%+(Qh;Qh75z>YnO)g zY!{|JA{aVVp38B3a1fmb$FW2xRB*c-ORn9!j%x_NodUVK1y+8L7u+eDJ0()+ofJqX zqTmBqF0=VO8aqwIkI882V(qKSD{BU>?v8b{*K00XUc*$`YJhEd8!%os&(^P9u=o?9 zjwtvv7>qNCaQk|F$FQkf?n&wC@mX_ngdtyPw!tbIPlajOHy-1M3Dd!dx@BXyTq>Z* z){cZ^j)@*ah2x-v1^;yWJQZ-e3Zw1`l5TZ@vGEHw7aPBh2$l_;EnHjhJDxtfLA#KF zpFzWd;%^W+jiEpE1Q)5$P3E~2Urp)+`)04+T$ITSR)B1WV{T|LaBa94>;mnA(^)F=nD@b@5W1$+jT}Y^Ig67Csz^?o zBG6MwxfOFM^M&UA5ObJ~$IE9)E&T}b35k*A8^MAdb5B65**%e>Mou<3(Q-@XespD% zQrAM&f8FMS$WL7bjy4)vTj++AEoa#-*e>K8B^k;q!QO~MTc4*75n)_5*OKykkx)Ig z95B|<#f_z*M2yuvM$mU~|9p2Rm&dH=y%ZAS3}eeCtc;IIcptm^sR!W{Z`pU*)uH18 zZ%W&;YaHE#Pn{E>Ed|3zp$iusM&tQSl(N=Ow=(05)|)(uTgkUmSrxD#D0$q6MS#wAakudtJac_LTlU#6q%djMDPYkLGtP5 zJM|ucps%Z1FiLWFQKRs?IA?$ex7>-UrgnE<^`>_Vj$HbUp0}i|0{JoS8x84xjh+RI zK4#M=0=ceFCiS+UXyN>>3&uLb)m%L4yjS4ypO_iLnW1yLrDoS_7j92AZX}1J;0kki z5Iw}>7j-#l0rqGicJO%xlTNJkMvN-H5K>|a*mTGv5TpolXED9CeijO`e3u``6v${I zFMXL-*B-2X_gWc$f?wCgUymMClt`BD%ett1^Yc9W02c$-Pi@%1v>FWP_O|z_x>&!fHrcUMKw(beoKO0 z(JW{);ykW;%}*IIBO(w)zb|+f2P(tzB69ljZm~Y+R+X1~BSj{7|1& zh!6op-O?bMYXvh_T8YJSbXrF_`h-HX<5(ndb(Xx3u zt$!qEwqm91p=a$(hSO-}_(HX2h~$p({+CeymN3kdJC>1Y`x_mo#n?}IN$)I^@2CzOXp z9@2Ud5Nka;>pc_}mX#%xi4Rki-@+4xOXFJdBqVG22UDzEC&Er5tK)m(tjU+~ z`;h4I1wP2yvYh!H>b!b$L8J3nftgK5*VetkW;auA*KHS`F8TGABRt!(;QbG%*+`p- z@S4xDST{1IJG8YvIY`#}=ov@j@cAD#r|N|Ou}~`V=TR8-rbF}w)K@S_ZoCyrQvzY6 zF{jZ33@SltKa8gL0PX~N#=4d$zi*v!XfC*z)Cy?rV!RC&2DdF&gk7**IHM?iJP(ed z0fFOwG>N}qv-lUl0F=aWcW{P`qaEXFQ^J&$g+ z-A;(>+mB*%vFg@F1zdMC^oB13>_4)o>@nH}^8;~%3spF{+B;t_IH?i@nzmSKMrkd@ zgRIV^hOnhM7+pLj>AiD#T#2w;x@2$4(a+-#gUvT28{UncsmbvexG>By@BCZOk`#UV zuqJ68`jZ}x3vuB)<}ef!Oe@|H=W4v-^ye;gyakWA&M-9>xiZ}qxPAL+FfbXWceS4E z%I$(FEB$UN8ty!ANT|rNYXA5z(TZ!L+1?5s*u=Q>3$T@t({6cd1H}* zr1-JuSIJct`9Txd6_6&k)7XI{d?=LE`Het^L)V|27fEF6qEKnpP3oiP;Wr{mnol$B z&V}Pu7i}BAd~V)O(R^$ zo)m7ri@JUGh%c0DUTglY)9b9YR~4H%#!O-fRD+{lG7&vJ$axgzeT9J{ROPSnnEl_{ zam_uFq~@&2sP4r#0JDb&Rj+&mAP319v(}Ht53!32_Qvz)Ot(N2R$))v2z)>5gZ>U}M1CYO&bmT82WYesLE_E=)kh z%WS;BtC3`L5i}A1!5D);7xwR_y$6(!D7nI3i>9{`dqP>NKt0wZ~ z>i(3|<3T*15uqpFU?o0R0G&gRd#K#L;XRO;-h3qK1NtT5250hSx!cC7jhgNW7E#+Q zYgeb*VrZHRL7nId2)@Mx7<5*M+RQD7iE0f@bMcr@ za|I-S!=xCP456DYBfCPo;CRkgC~WHZIMDpU!Os_O=^c1BJorH>?i3X}gdJk*vEvB2UjH zhkHT{erQ666yPp5hCY#8$up4Cd|RM5AW6xW#!vZad;M1 zg*Dz{&-}^rf$`PnPcxw$i|8d~d z$1kJX7~YDcnG?OZJlZQ*wfsiHH8VEMNi-tC;;ZaA6?*64LwNJA&Ur||Z0X{C znl4+qq0FgdR)rX$Oj04#r*am;+q{K=_ERYS(}+WyzpaAhW0)pL`3pG6DVoBt30(w1 zPjV`kf)tV*?cR^cyrQZjN48{y7>Y}sIDk1i9fGX%-y-O6((o{i#L)vJ(GZKnN$1KP z_&DiNpn%-NugQ0cO%6tNcbWMMY&>qwMItMp0)uZjYyT}|!}C8ko;pLMq50M8(Ym$^ z%p?}Up&p^Xnav5lQIuvvC2vWCm*2UM2cNE-b^Q3U27BJ{ojU|-OYx1eKc$wca7Cr) zzyw|?G2AI`pLC%SjL*|mb{4|w6Otl{ynu-*s zICdZxd=H7gI8={Acm_CZE=dQ|Z>}=&rHWRXOYPWCvUEX%a}4Y_tgV%_qGF=s``T#|n>|=_ z@#5a2!Yr;2mMt<{%;lqP?b@{%C8$i%kH0c!WSqV%jAuduFKMml_IuTEoj_)JKn2_J zmv}VvKKk>2`T*k@l@_<@a~?*ap!f=em%`!{zlS#7ONK%vlcU)*itug9MiI%&@i?5# z_`@Z}rS5ic!*_4)f@k(BR{L8IzahgA^qLi7z1oGRzhcmIo!PN0kg9=H=ME(SzVbZg zGGARw6*3yDH$#$MPs1C8&I_2j22*k`hbYnXlDeCmntmdZ7>$dRR&mo0w#hTt?Dd+9 z7-pyfFGzqEIy=M>#vhMRA0s zRoC>?8Ib0pNv5>|PD(l5)rcEH3|u=}BzEC;L60(nk|INlIn8c}IE!o~49E~wO4hG4 zPo>!N3X|5<{dPe*U&P%}GQ%Z~TP~%Rn^uqi$ziDX!@cKjuR!Zl&sxCenH%~Yl8Dh0 zs$vUVANs`*o#r%-h$tVU+68VwQUs~_C?xx4 z6iNcj0*c2x8LQim+#%|cI9>T+_v#5XwgpKm=?%EaHy;e*{K#Ns?T+$Z0N z^8+GBo+Y3O09pa$-iO$($Ss&c>qMpDVbtavz)9~y8s9QUucoR!NH}ZxgL7er$BDb9)C|#hb|i>iq)Ab* z^H`LP^l21omN zkLr&~bK%MqSD^JRM&BS~=-hI>*ah2#LGeZgVtPF4E~D-^K_DX6h<}K~a`?G&xX+pw zq6~6Xl27xXg79W><~e}5IwFOvaS1`VKZ?aeS@WhgVY1dsH5WZPJr(%bexPk}8(K=I z*P68p1H_M6|EF<5WBSu9ns&b?>3Eoz;VE@=t_y&zx=_PFD)+!z>N@G@Qd*=`q|$rq zvj55+u!s1GM}bf97Fszgy06w;FfdIOP}pEdZ86)L-*7JN0_{Qrzq~4SdWR6Z zW{R;RlT*+F`SK}o=n5*KbS9a zc9!F(8gB<=Wefj9iVW9w>Cg>}&e72{~j))2Az#Mw34weP&$ zTLWq^f2(U+>{&J#m*#@$b@kiqop(6?1{=fM)+@&@*)AkXgk~v@fPNhG?~xQHit6); z9XU{?-+XC`azj<@mP_rgSF{j9uPA6NV)Ch0>?oli85Pl$g3DQ-1*_J_V~{_mE)sq? z)W@bDTv!Q@!P=))An8huMptn(jcXr`@>|y&qUPe0DX-Y}?P$~uH!CO>mf3FBi`}JN z=#(6fF;>zYW>NgU?aorjg9AJgM_D{XHaL1Y>!i>gNqyH|2})n1@P52pt2CFZioaMP z!ylLZ6FIC!0P{LS71<}IobW@U+Wjk1y*Q=KbKz@2UyaC{tevx1e@vQ-L^r<`VBGoW z8(>zjTwh%^R*PM+T@WPxxmc2B96i8u0)P8Ni9As$NN%ZLqtu8OEaS5dQk^E}Qm9RQ zpAE+GFoMVHwz3yv@eC3!f@2YBA*jcMAaC^OxPd3rk?F16VJG#71aZ&U%fRVy1pTjE8~tSyN*ongCuyAUSIO-g0D{BZCg zxJ+)c`6Bpm_PHEW(&!yaUs`o&MG}5SRbr8ritb7Xgcu;<#ryaYQjUE5)9n=xvoyO8 zAMp`hqY$+S@9U>UzWNoKiwXv+YCE{mINC}zpuA$*tW&#|x9nGzvOW;AUFA5Mz@TUD zmAX6f(+~+GCP%0QJ`oWW))1O6EbRR>na{fj9Euo)*E=kA5Ap;Wd)4RBEu2IS#hI~6 z-n!)~2X12nT?0DD>CkCz^tP{IvyTAvD zV2MXpd<4NE#bK5xrU}Qq+j|V*sC$%@TN-T-^j7T)NtBK&;>*C4$8eo%!aalH$7vH@ zXU2EXU)7J<^t;c}Du`&;#aG>zZZ5!d{wff*H%d%H}#FeHl>$>(@@NYREs z{}1Qh!bH!?(UW(0hkPKl9tu4=@atCfIo0KW6`CaM-)5twQ(Q82>>|b7A?STdVzDA> zUjkm=M&jVt&$#K8)LNS-zWLiW7dHAX6|i~rVA{$yfb4F1?b_`^4bl<82|Ft%662#$ z!NoE~JFeB?^~;Rxo2 zR8UdCXl4bDIjiH~opfJGmr_PWJ$330U%$B+(mAZa&~As@z+gbDy(sKb?ZOp1CM}9` zK@P=1LLG+V@F^llHAx=G(l0$mNi$pkPhy3%8O2P-Gf7|af}YO#wvUkB+^pJv6+%$l zO}qqpYlV$R$@L01-$ZuWdlsZ1`*nt|!Ew9&_GF8|@P=>QTsXCS5MA~Td+$$!!}q5j{&E@c-3bbs<6wU|7)M0{xhRo?~LcYce!?b8({hBcjj2T5+f(NI>8$d2%O4_X}`cJ3DLhH*PK{ z-NaRFXzLEHt#*UXmQ!vQY!`fRQ%EUT@)(ihz3!cZn}g*P9}C&MFPsKxCB?X{8Cmkl zWbR0h=HWlY&Eh=#2&WBWEqhx)jrvXe-4$RUg!Fl`8E;0Sh3Kin;bev`+ou6rQ+%Gn zcIY5P*^nU61QNqGuM3zjER%$%ccI45;R zK(nBRWqtNwt=DWWd}O*R;PdNowAF0L*wK92h1&%aQObB z1+H9+-#0A=CPV243(7jT3yFuB6dHHb4z0igFyVJ-xgH4~3l@(gpj?i~RhLz~nJ&m0 z(M8M*coK@!+Cj;=@!fqDNh*SO?mdy7qJ14Bq2Zh$dt${PJNb|NOK?Wg*cO3+E#A7h zfO{RiV>bhDXfPS zw+qIu-!3ffNqNU)8eGO?RHehO(G4-UG#8-1m@fy2$d8tkTCONHfe21RjVlqb4fArQkBoPQ>Xwyx;$!U$X!S;|X-XhVnzy`S?d&`s3Ea*jy+V z=b#GE#g@NewhWVo`CYC=yL!72uda4{E^=&)@y-v|tD<*rMT(yNtRCVd*{t+v`9eq8 z)oMqvw@M<)Qv#emf`sEc+Lw4i7I`MjDVYRO(ICsyIdv(wt368+l&Cm%9=kU>DWzhvO%wGAg3@-QpSw;19`&*7-#iy?(YjGaO_mr>!LUhAe*q>nN!}I~u3KQ+u zYCsaVyLQ`UQP7vO=yc*?dBho%9W^@wIVdY`)z^Ma!HMOWA?S{VSDYYCJ*7Z!y&$4;JX*zgf5cUkB7tfZ?U`Kn40f(H^&>kWplCi zI-k+2$I}+I;bqIowF|ZjCF*wxj-Iw@M45xP%j1$cdqU zZ^qFNY+$eU@p7{fXfA+c4l6d9yBK1_fPrd9i@`43E^tX;U_suHkRr(YA=O*7Q4iw? zIqto2GI+roYfU@aw4+Uh$)fhBHyB9nkdqNU3{g#}Odrx+pnniWlBZ=JIP#Y>Q#f+a z{SneyT{~3cRhtVXnXd{cZ8)U1xD7SiPOx3EU2u`pi|BPq`aK%h<8BWZij+d)@O4S( zn<(Hw3qqG0rd1c`A1QRPBn8FLK_n4Hafv$rDZZ6~cvA~II8F2S&*TnN_(eWi2_u+* z`Wdfky_C8wUbnf>d5v_!E5_7TxM5}ovuhV^7i>Pp@Qt(JlQ`@mpZ()Zf;by>U6%vK z_2@W4m^UJM59d^rLAThsR38G5lC+w;9t(KkAp{=^6~`&Y)HvWsp_Fr67W(NVrf45) zV71xvZU{J#c^N?0UlvODm~y?f<9@Zqrn%^nd9E-_Z#e@79z*2z%g3(SESv;rBj zsmt7ruL-OIB>l51uOr&0`tcHngoq7=V6mUZ_mZ2Zf6(;2u;Y{UNYnfBN9V-F^iIM| z7-=8t!*QdLX)X+z*a{T>lM`czGL*h{ZCTHDVX+^7r}o!AwA!B}QhhF7Kbwee?y16^ z|5I0=N>J$PYxR8*Z@%~Kzo!t=5H0fq!hrl&@dXP2B6Xak1>4C3f`aJ5IwiU3U5okKSyaMzvFSwjQ14VpOKQ0;F3GxS_>x_9kn@ZqqJMvZaeg z-NP(6&4Q~0<*p%wHQeCcJQ#*95q7V9*XlYzH7j;d+*~AcjjNGwOUAo+TVSi*$&Qt} ziL+$RZnkvmD*6ujL+OPM_(EjuZhd&o*WS304e4q9qP6)wHh%Hu;*2S*fZ3)4YYX0R zwWH~_3%3hN2V}-Y(?F`BFLPqqpIL;IB)qfUbp9BTQt(0!gXi<)&0%u+<{e5yhp;mj zq3t};ADZ76ESBCwnWnDRzhFXlg7mklrFXsmhm!`LPTN2fU5xG7r`SnnMEjH9zd}p3PC9Tq?vQpUkQ}1v+h^9tN;;F zNsQ9|DDIv}1+==BDZl3{U%I(iV$v#bwC(WPk~ds!HT8D6b|L62K@RRV3_eh-1$kfq zMCHwtq)z+*M2Fqpj(_Io`*KjN_rf)h_cE`5Mu(4p(oE}EJ;aM8%?6fSqh>1dsyv0A zpMNeQTkb(~CmlIBpz8>L)`ijbZ{A#F<+E7EHH8vZc3IzC4peG zex{fOT;G$3b0ImQ`qfFC-jF~_4X7w}0q;FTPb7Erqd)9jh3R)mc#$aBQ2U@qUSI?7 zhs4j(U*X%mln{7bV_R<^n~QB-5f$KW(3BZ84f9*BO}k*buzpJLIJgSO;+}bbdm9gM zz*B6S!h@XOkT4?W`~_(*Yu)j!JFYcWc`~D5nsU2fMs`qA-wWp;UCQ~gNsf#JR=~S^ z2yti9N9-6Cgs-h65{Y6o8$IJqFFWX{5Rq{$>!=nDO9sB;KlAI z8PxtE9B!h~0?YY9o=U&i3n^ZZanvQXnvJWdrca}Ys@vzzWS)mPL?T zJ=c~LGHDC&d}Ci~d@cocRom57)%~!TcX~6)MRNuXO(UyH@2{liRZsaUAT)nG=pgdA zip70$HVJCSxhD#I>9DSOt2r{w#b7>)RS?YE)Ea4nh@tFF zmx$f2U5Gru2lAMR|0nYDuzQ@+hQ6d?xdHr`#E_D;9lJfXBM7BS77lG@J#wGllSZY2e(oTX|A3R;pj9 zfe1I!VjdtchN;_La?FC}h_06*tNi#7189HBzGz;GBT;p_JYz zIGHAo}wmPPyw&7V+n^o6jZTMcz1&vNlg~fXHNZJB6K-8Q_yHLB3a7rl} zRVKPP>*DmA2M4hP$0z8iD#N&_g-LXr;37sE zK?^jJ;qpw}(aCx6)QYkr!IPwXN=Y3@?UH;@VMwkb=QN?|E$g`L z2sRglx-Kd}_;zz)U^CRd@fx!mwhP_8NsQ}197cDPpqU2yy-N+8Q#d$I9}@iQi@<9h zyu^FilPoe&0M9;0)oLt~9$syC91jgPo`{id|Va@XH>RLj>nUW2NAEDfPDa8gY0cMUz}S zmTN5P{6b!JOU$rb666-CYO*nPkKqst=%~x2|8~~(+Og91!tddCvA+Vx@Vk%feWW?t zv*{f_FSXBw^+%z(Q2aH?dPjq9$S`cxT^4qwc44eHz57mB;Dq?{Q}~Dgvaj!rjKU(K zxWKY25~(rFli)0(SXL?2@=l$fI+0n5(?qtm)A=IJv=n0pF~^CQk=#$lW3*aH)hpsI ze;$5?^#K)eSJ`+DngA6J&6|-A**({W)TU?rau8hj7%w6LWZgsL)}zy02<9VPWeoE+ zJDLT9kb!Y03&$?oE@a|yyc;bkFpS({us_bEo`{&f69UlDxWYZNYO6NQ>lq`5;Co8p zA0dIUSDbU)i)L(CQJH)}E`htN_fwxmv2`BeB<)MGP_E#$n>byq7j7=VWZEiDjGK?E zt#gCV)-!FFY!`GOa0I8}BfCVPk_tcC>l6bYA0;R^WY(c9u7Fa)1&2kLk|)t}Cb99N z&Ek%DF@4g4WUK1dv99T<^$yKNgG@&S8p_Acmgj24=QH}s39>7+Yk{9c-w9(3xfeO}_uWf6}Z@xsroVNY`ILY$bZ_oE>j^pU{Q}*dQC{ZYd`*aaSPOI)nyUpDDcQAT$^pmGV_m3Z&-H+VVF5?R*v3 zE!u@JG!thCe}vRI3Pe9~9!{1)KN{m=3%_T<{wN{R6*u3bmk=`Jgo zBA|~_Tslk8^%f}j$}tv~?kq;6q>>5lF5CSup(*(;gJMNDdlcM)t4HlzI?6D9hy``q z!r^Ecl5K+q-D32B;+!KKzaNiyctlwxffb};(bRuP;kpT-?yGq0>sIVhrI;@cMrroo za4k$`c7F$?^yn8#IignaZG9N8J$~JhZZ5KInyQSSHEPUU8mR4eW0woruHLSnj6(oD z!f$d%e3P!BZXaEx}f^u6ti>@ zBhHZ2UFMs-L~~cTdy9${zWIRKqP2xyJ%84|U5orO{I0I!;5@t|X-tZ*s7`5KlBIm9 zYmF9JqQkR{lxZgNFO|m{rR;i6 z!%W#0g0^old1f45z)Ekeju<=&q_`nHsnlCt+-x-z&4o}tX`(4j1a$#>LxK2eg+VU;MX=)HAkYkU~_X? zVXt;Q@&*w@RsAJm*J>BsR9SW`^c>DIgjk^>gg5Gzg8`_*jfJXW<-jZSa?#X6uOijA zwER*5fjo(5Z17O}WC|w{R$tX3b5zrB@hvdaKfC(Se>#WU)|JXu_pUiA%>|*%as>o; zHuQ!U!`=>7iCwf^Ap9#Ek~2jpYw94kR#FAR5l5#%IqsCVu+~^4>}hOpD}aitOP-|M zNfAS-GroWc|F;8;WC;JO^CuE>o=jgv+}41Q)JG$FsX zfYrJnovY|<&TlBfDf$Qso;}n%?nJQdDnT5>$#5J?rV4_^e0woM{%tnXrYERD0j$gD zCK?Q!L+|sLcOimw#%(jsfc460)*X`Of+`>DDrDtt-flh&L58_a7m!_{T^PIy7u}C= z@zBYh28b8J#tUv^Cu74mrL0)jtmB)B=X;>iYZM2gz88nGcbzDlFFTyfh`bK}t6;L^ zMnKHnR>7oItSG8C%lCZ*h~?*iXb`S(3bxdP(xz|LTnOphR2Zg>ht$@xp<&Z0wJWp> z4JXM6hTZ3gWgGrfi72nJ@e(ECc?FwQdv#KfTm>z)H~2dPBT70#Q|9yZ9xZ@rp5k9X zihB&Apz=_Pl&k1FUOy`FbuTFUEE!f>o+x(yiqAx5?>SLjlg>~0>E78TELbRlrodsI zEo&3TYmZEGVdbZ}0$6V{0|p%f-wqazU9??zJiuRo7BFl(T!iB>bc5i-Bt$pkY^|!1 z<06z=Ds9Y`?0E?!BCP`yo}@7JT0S#c;ROJ$-xq3-Jd48cTX@e8MDjq%BN9pUtIuOH zq2!x`Ni2~t@!g*yJBo_%J#3^evCs?aTC#0Mp}Fu|AKj|W;d;AW%z{D1@VD_Qu}icI zZM_?mZo}!4>Mq$0rTAdzvC7p&k8Va6q5}UJ=610ZaKa;hj3MEYba=aR&>yb#2MiQL zA4{?^VK?%xp!`UQy*09g-X7zYKaaxcD4U7A=%>X!GB%DdaU^>C7v4BtQr0ZwoJMF_ zQnLhshZ1}nOM)@z)Njs1$(}lg>lcsf!qFb9-Y_*6k#5#2?BVa37eki;wDzj9OSKEA zd&wPz`GP%|6VZc6OWnj%f$6^=AbU6@d|ir7YuEYr1NgDD6O=B5!TX+Uk~?R$ECob2 ztuCpgQ>t|1ZjaWDK8?cbBwfx&BFJ_RFm-RLU++a@v5l7rzbcP+4WTNYSH|w0TQ0R4)8W3Hf=KXIY6= z$%jr4%(?%jQQRofC>VJ1foHU2glrWYA6$MYw^JGoftu+WcBG-?mOM~SGr~7N zl{yYHRaA3%T)CeGshZwUT9M1~Qqr$W>u^t{&^mquV?BJ0M~}Wt3%TJytDoKD7p;q| zjo-Vu79^?kkedh((cM$0_0xDR`SN6NEJo3zeucUd--t?|R}1~B z(dMt1CQ5Gldg_wncbHKTtuB=Er+D?7Hy27ecNLqyw-|QAg8^=nRbjo`g(hOOM0Lfl z2#a7K+NUQ zo)d{Ff^O|;G>XgF69mDOJ zamJITSG>`UDy_j48P7&_!MM%9H5Zp_BVV;qTx<5`6J)?M*zRi0*_GP`>G#7VBi0;z zT0+SeT^X+fBW@8aR`6j;ZLo4QuQeq{ncCFS)P?NM&wagj`Tj6Y=Zo&Al-~@>CSeGj z_!_d5UvVc6{N)S*qZXE?uQ{4EsrShG{TGI!4z?hYW$HO z2cu42xno~mjaPjz)aTUZyODv~dR&@|N}22mMDB9*4KxPF?U#yOvt2-x(hdk>jJ%Ii zaxNuyMx6@a84sg-Azxl_dK)Z%agxkiqpc&YiyK{$`NsF+!+P(3%*L8)@GPdces}Dg zf3(iYABr$yi^BPW>>|OvsP<_wzwUOg++3iok75<#*4zDhv~B$xrs_?&U8h|*+KZ;& zL$PV^2Sk_`uT7agK03R6*Ex=(R4V=v=N@;C!i<t z0EmRnGyJ%F6K!*VIf>Ds^58iMv&9n;gFosX>uS7Ub0Oj)ToDD^WGySJ9$ek~_ym@nXAc zIu3#GhBhG+Cy!q zFre*hE!d^o1-Jv~jBq^m5$;Hn;1JDeEE1uCH58PCOq#-p8bnx`K_VJ8l0c1Uu989K z9r(!Z2(?(g>+~ptw2ahPQHM57!of(Iny@h$rFcnl{<7~!B^T%8{5+jY4$cEqFdUVn z;1qDVpF}9Nb%xNy^d+G6{r}y)?Q+}5wx;=e_En(k$JsM+HtjRJ&NZ_qJTm`QaOQO<9VnrF#Xh)iN|ftzJbIb`~8M`%GR*^ zGz)CIhaXOlKeSK7&xh|mU~z0+rC*VYH^K;yta}__uBn{9d{b{)?8OLKIj|YEm7?Tb zw+l){dOlLX)@4@j_OsC-PrY{PC%JZ5D=XJbsq}*NMUG5U8Ek$ujkb*q4qI4AyJ@qq zA@PGZ1Y(3j3b3&EiE<_!hr>q<@uu)eL!mevI-|uYm1nOo%cI^|>GXn<=eP&*rz&jD z#KMQweVPY0>ZxH5wgP_y>u??;6ZEU+{dARx~XLBbV_|zEP|Kk6WECFqyz>~{W&sNpm zzP_lENiCB^Tb*=6gyCuZEn)X)7OH+nKwW0eprJG#Efg{zh94eecG1W~)j!ybX_ z5l}gu?(&=o zx@aZPdh(Ro1~;s1Wqs|o&BD#8@>UZ}F}M=KDLlkPKj7t}%6guODHld1vBZlWuY6Lb zUU6OE(}_4=b(K|zL`wpeDYy{qQe9aKsax!!KAPdaEf0P$7;-* z^r))r1l=iIAve^>RDQ;xYnHze$!9v38gr(_obv#`o#ucJ=A3&CyHGb3#Am8tSpkxg|d)WWOwXr+tk78HI-)GX_cy z45M&TEaC$u5dClnl?eg-2NNww8|+mM;8^D!>kBDXsYHZ5eHLva8%(ycrgqzA;e+_x za4)%nW@ju?q;&yp4FP;Om^tc<#L=7zi$;kEd(e@+t{pjRSx1U~%>;5j;K`LsE@J>; zKrjgS#IpTpK97iSqNo>f!%jePs*;*sABmf~)hj1azn28<$v6CLG<~5LsZTbIufIjT z7>)8g>LU{M`Ti#IfrhNY=E|($G0Kc6r zY4>dw;xVhYV9gMWb1L2?UW465p5@$D4lZ4@iBI*<6AcDeE#O z5=3f*NrXoi_(8@4ffuUlvg_M^6$wQ2Vf)>u z*Kfk#-?ZO}Xr7S1O#<4tI9wt<;QUF=%t8E!FK?-v;Ap91karYYCt$_1tv{h^F{Q|p z#N-qA|7IkKZmJ46S4E6KokJ@d?nU>N(Y)z2))&}M^I&X$B@Ax{?Vq#F?77Xt`Qe99{Bo^* ziBq`sW*TaFe_{kyGgT9b^^pg zHa3S_@hD#0r>z~KO_-q+4++P7On^%3V9^@wR$Hqu6xU*gOqNWzhuJ$kN-`00F&0fY zB9j&SCBmI?Y5|%2gnN~iveOEt1CFmS=RsHrqqAN)Li%>~N;>c^$yk|Zm0iy@$so&( zZiOHEC?R@{zk<`swxtSFRbL#+l9y<|HCQGCoS{~?=o2@c^==kiwRRIqphN&rFYX?B zl+MCKNg2h@B4Tazfi-U$XjAM*DgnIf7|Ct#%0hs4$${=Uaihr$Q9LfpWav@(18X0KddH2F0w0o2A*ZXhrqt+D#n zxmgGzB8z&}H$-?+8vZ^FqgnV)i1}8Y$!HKnvp^awmoUwRjiSKOR}Oq23>xK_NVS<{ z(ur_-vvbV-M^cKg;gDxf=v6^sJb0<0F$doIN+e8cmX=O^wC%<=n5hF`h z0+)@Z)pol=w5U^3Kb8yJ(Ro$dUI9kxDVOnvjX zALJl5KmuXJl-(}|5AUh??pWq*9*jd<`dYQS!mfN9ISYr?2^R98f`vn(tYPMJ*Y=@| zMh*O1@QK*8BpL=6FZ7ld#Bz_W2v$>d5Gx4;4plu7Yt2}FVaSS?pzzb5vLuEvgYR~B zo87lr*u^F<*d<_NSp#&nd0MlyrJS;%MyHhp4lH+eR1VfzN)C*suCsF_c4iRT|#ju>6An>9j++8Z{-6 zLR5gNFy6xC)fe19j^TNRl`xbUdY^A+*|Qpk=(HaVhnPaiVd&4pW!9p+Tyg%3xLP_d z_YjlHa?N7J64cSKAOR%tNc??IuOX}t;&kX*Yq-#q>RY#=(Ksw6cwU@5^ zNKoH6OS`SZ%+Y3kj;a2GvF5YbQhgNAM@iQB#^W7WZI>IZ24*>!n1$^0P~EFQWOV8yWXn`}px! z1Gz1-Qkqhs%bErqpM$j&3QPLpbPT_>Fqb_De+^t^hO-pHCs#-r@%jMLerSMFMyfvt z6uQG_ago4>tDnUpM2hiEg!bGa#|nFIWiwEv`KT{C^Mx(JXT=FNJQ(DjaYxw0ngzZ? z4v<}}4VYp=3Tm}JrITK(eIHGvfWu5<&yj+w@mFl0hsY`J&g1KNG7%X{+<#6~p}O|t z0XS+6^5aoZi8g$QLMw#vPOCl9h#UeAIKXNXIfK3DF;~^y_Z26gzJQaZECJhQ(`+y> z%x%64?4Henmd3?klu5(A`6wDDUEvn&DU})0EkFOQhL-OkL8y6;9AD0mWPYMt*DahQ z<*Ab7>tW+Q%8^h(gNIPv86^C&`Sp@e3#uu0+OJG0SoMADizHpV5-4pn&9>_eN$W4V z-J@B^IZ2Y+C~POQSu}_!9tpdIuco|Ua9NETZR@!<7j1(NnuCZMILutN6yihgr1%!u z^+0S|PJA;+|y8KN35ZibN|?9#xoc$zG>li> zcy^~|Vfn4t`Y=X>oo+wHHzz58lLey5gW4zPat#6)=5;2 zgjbS{3_~q~G(-su+c!{h-XIP6yfVfnZ>>mJYMpDFOBTyJN#6fK8iWh*^yATczJohM z&T)L>^O41bLYYhb_IGS0(z>dQtoY{j#hfl-iIQ4%f^Fv;#5P}fyJy4DB$dL!i!iWZ&X7 zh#;rArrwD};1oogj7L;T9i5`*GQCab2(;5E<#$q|bp%glQ~BKIlfYj-Fder9cZ1{_o zDzo0eo^eAW&jfQlDZRP51!h4M^9@YtB@ot^ibLEF>glkW#M8hh(o7m%C?!2yer3 zSht(+2fc-R6noB90;3x|p-4@sSE&99o+zA%Cl5(TWBBZ3-S`kC;~1|-F%_Q5x58`By>Z$BgAMNewbVl z472josm;V(3OKbHe+&J+mLwAa>FlZ>JH(X23sY2e6@NsQ$wM8~6sF`9BGC21foQNC zAiYCgh!$`!kLH!#@&0K~eZ~4>OxCCbVjE1YZEnNQx{GXgXcj!M&3_XC_gW&46V%KR zOcQ9CL|q?4=-cXuVMW3ziIjE z4tnAwxSQQe??_6IT`Nb@O(hp|VHoL)PD~oHl zZ5oDVGqm{q=;~?$uWuYpq^3!T*UnYwdPK_jc2dF8wbT7gjuA&Ep=%M{P^^zy*h@HG zcGBJ)@0-RiOJ)$QCNmfg1420mlyg|B$Ku8!$dRGxZH_UMfE~U`+=X4Du%!;P`0kP+ z1t_%H8G$(8q@mhaoag;x{LL|#o~O7by-{>G>ykU6-OnL{sSd_!N+U3pZ1AJD9f?^b*u4^Ss0>^+?z46>hl) z{U}-H`l;jZP8dp$myA1#!xEQ;j7j|Mj8B`MrxbPR!0+ONXZXyqHNEfkxdf1h$t*Z} zQ5B$@%|m^`sw-Op)oQbBU@*9CaTnOln}xK4WDxbl!wzMVS|M(G^d z9yI{`aGBJRr-y6n+^h}mDrN~eOG>^?ZyIU@7O4S3ezaS%jK`Sf zh{eQgAS5irS0%9D9!#QMLg}!=CU<3Bd8!i+W+q&u*Bq-y``*WkxY~Sglgh=5)_lSG z?tPiLG807QnYEp5uvmA6?GDYt#wDD~*w&9po;fFEP!3Mm$~Y5DLP_9!1Rbs=5yRTj zrFzO)>Ge(88ahV3h5YcDpGwiv65E~DlYqy8{l0W&VhGSE=x`J_M6S2m6!#;7`P~qY zx7>dR<2b-WV)Q1!uByOT`=#rP7+JUySy5q9ZKE%svZdv=J2wnMl!Ba5FDraAqGlFW zy!?bGCJ4B@W*Ir2D^DnSLpciihLDb+JQ14_4|>yJhMOq6!SaTeN=y)KtbKAjPcmD; zkAJ#SQeO!2 zs*JZvGgM#D*F3Bh*T?cn7mc8&?B%NYyH8k^v zfi(1>6o)XMUU{s_7fMMa3Ly4?D=NZXM_rBd7Xd2YtC77u3f@hKUe-E&$_`xst{)r0 zSI#iVoyJisngYA_WaebyG#ZI?1lRhx(@$_xbLAxI6E5kl5lr$o$;g{j_GGO)QT0V= zzTPDW-Tt&2d<>A!yNm33&4R1M8WPq7BP~_P4YL z(mIMah`OpU3-yH~lUSyuZaT*X1OwjoHiO-}Ss-)4NZ8l&_@q#{d(sgB`~I(z^Pe?+ z1yhgo`i|^JfPf*8*kF-|OO=3}M1=A(oy|B2ei%zM<7C#-8bcTcjrz;-U{_cX7o*l5 zc~q4KC|t}3zEsfakf4KvdbtT{ah>s;nz;_LB-ip5lITT9W*#VVl1;{a4EP!oT~+n1 zJkRw7^J4bPFuiE?)7HqKYCi*i!bY`0H0$?53OGp-u|wE~P(4Rji>@-*BG5yL>6$X* zc%xCNSDtfS+d$g}*4F(L470&3k^ERP8>lHZQKM@y4kdr42wok9g4Sgx2*~wghd@vy z@D4@A0P;DME0}uq8+dcniu8JzIHtd&3E3A3txlvB`APC?rD6HIPRM`8Nw390ht+EY zR-Ul>0zY4eGT7&B{*<*c?AZ^%`a93=(JcI)OLkrls=tUL*N%H7JWVwrJ(`4rI0DNn zvAb+jfbkDvKkFvBfKyJsl=*~t%q7*PqqM6DN*zwXF-$E>7G8Zg*wmpY=ynEaJnaIt zWhCyBPBcpIr?)89Z++0_aZDMdQGBDtvAOlNzb;%9D!JP(CFyNtKYOJKsxMaELYEwW zf68JQm<*7cZ6~{5v%sm0DnhZ4NqiTs_z4`EQha-Y2j*Q**>LBp;py5+Wlq`#d0Qyw zsk7@HI=P#sfW`b_p3X(lxEDhhgFBwmViXd~oNoMq^4hl;mOkRKpToX=r%gLEl95of zMP;72bMNjXhHeJMnq*$A*HsnfigQt4B<3?+CN=Xm&$b!{90TGS8^<~~3w0f+n{FY) z9fg;8`>~tO;B#z;pW{jDiizz{FxDzF=Ss=g4y-A(WWK3OA);zvw;xphTL!>qSqQ_8fv(z?vHLU&f0vn*h$U@C2wph8WnFP?_QDReQep&O zPK&z|e2ex(Piy-ExyyMR4O$ntEG3j%Vzf1`n*(YAe)?#!-pW7}b?#7koaE(SB;Da{ zGWsJo?k<>0w7FyR6tZ05gV9S8+H;s8%B$zXSDu*qqR=gJi8}hsOJNW)C|2D-cBf{6 zk?6CqBit~Zc;E=!mjoeag(d!$)rzR~Tat>!e%=|H-M|(zlVFA!BJqlIUD4VjqY%(J zQw;U~Fw}^Z=LTDiPof(V6=zt zx~i7C(R9@prn&=4Q2JBX#*ntaY`!-u>@mArvv7ME4Ft93D+N@b3(se6)Nx$2W^TiyV)`MTRZn+2ox=vE^Rv<9vn>%wBK_zF2_-CKST%<}qOJ?Zh8fAX-Y!=;0Cv-fsW0ICT9*KB zyDMPGF_dk-VeFpGLgQ&N#6TE@7vV{dFBKb9;D!#T_@@1fumvD96OuFC6dE^1r}a zq)j!wgAlV+s9Y#gE1xw`^?|68XUVBZp6v9)Zl7qDTci)6Ha?wjKbrK(IYjgG%+rTy z62lv|hr8Wz{vE}`t3G~l(KM4u*6Z~vvqMyvh591Mtz`**w88T%fT6_DwdPK-8#D`M zyUAGi??Tr2h!0Aixze8Dk+R&{+3wGG_KAiD zsmf4h2Qt`d$KoJ6KRqpIfjn89Yo7CRUn(w-dYrO|^~SuRI6B1ZK8W#e`+&`=x?hDi ztuKCL#Y)6fg=w|@ZJ^m;jqPU50?V7*xDT0mxO;lqI_u~^8o*I7U97~5>p@x|JzdPl zvf07~LaH_0v$#nFWlLyh`Z7^f?h1t`*z?)9FVaSmAv-XcYxmoBv?0jgSw%&w?@4^nHAIWn2 z##Rsa6RNtlTqcxDrG3P9vD69viA5)Gt8lc-7&b_&ad(^kmApTt!z7hfXr>C<-{Cvp z=7ywLBqJ8UeToyv@qEq>_WIGS6wxB!gA`g?@Ml#hZu0u|#UhJYBJHZqu_0iAy~U1s z;^O~Hgh7i6K!qJ(H*OZRcGB>7&P+r{jiNEfs2dn07K!DnZ4=TI3mi?CbI_$U*02Bh zZ~yaukoG!CaW6$h)lVNIB|8)|(2Hec+hH#!WtG5=s@Pa}{Po3yTZR%yluw}T&&9@n z>M~i^W^HZAya~z)FDX#)l~c<|%ps@faE`=C*F zwr<(*K8i>v*YsjP89!=*V<6&(l|ea;PU7~^>6}8RS(-cE6n=?3Ipy1a?XNHSwVdvW9kfxEIU8g zaVafDx#h6uN21WnBf67O&)NEwIqT&<=)H_a_pt-7A{*^c`%1M(rxPCVM^%(o zoR|7yRVKL%tk1a!h8@G^CVR*3)+{7mCWQ9EEB1zJ1YN1^_D#vgM5eYM2_9T-C~m7) z=#0VZYXUQJTCtl+H12!D;)6Uht&4+_IPM#IADBl3bZ%4%tnUjKJM!5b%1L@Ni5Tq# ze`?bA!Mg%RWT#U~i(*=*llz2Bw7qytskxQ5|5Y71+PRLX#dIVgdWPY1rY-v|@8kqgA1|>3i1~i@KI2+HCoJ+BP>RZGCm^ z5zRu95K|KGi$#M-4v1>QH-kZjq1}0!5hAO><4A3n+v+ZNp)u5XVbbc0Jh%8I(EC%C#^7d%-o}=*yEY5N5VWVQLo%}Ql!Yfk z%hkHWs4pzmr|DfhSdHODLkVlth7?GM?_UL-`DjLYM)g7jBe{cqF?F#dsNTVN0nX~O%ZH^I+v{;u5ev&%_&U42@gfJobQQ0rN!8l$Wn_%DpyOHR&WYg3Q&p^vq|j zabS&As1_o-9c8ZP@jQ=PZ@}S95(89t2iRIcF@!=7O59x{5oKTwgyN&n1nfCL)JvCe z>5{)AklcMyRa0I2&FhP}%eV}1Yt5!@bHmXyEVezQSy+0HOhC?shgk?Ei1TycPk@@8 z`f^fwVKtiawt;IaZfq&j;8UfiS`vuD;YrJXFEz?7;J+Z#q~5R}gqL5E?&)#ll+&|V zIPD6e>@vw2@5F6^HAXBnRZ&%84(f|57P3T;ZE>~@5Qe&qHig};S@`-u+)4Q6k1MKc z9m=`S6%itPW$4a=D~%}95>Wwx-BQ&csj_uUksypCe}C~Gd!%bdvmq|^^Z14X;S@7X z0ayCWzXkGq#nnkVfG`BL6Co7c4cw`1XA)H#S_K87}yid+`{S+^}b~=i(**Cyd z9b0R^e|?eVR11G4ftEkC!pzhcembFLvTYkPZcs7IZGW5CGnxfQ zu{MOi1HJGZ@s$o0imyjgvM-4avjf*7!Il~@v*;M)9X*Y*utjzqQbBwHX!1xRD|0|m zco+hN1SZyH&hs<4+M#0QTmYeLRajMj!}_917pz34tv|cAy$wQZt*v!$7CKJSenbp% zZxVll;xQduM;yB*38Lv;Xx?p{1*5xkkcr?zVO*&j(&5#TKfKT*mqaSP-YJ>63sw$3 zjG+9N@*|YpD#B+m7VH;GgH(YfF)kPwj|j^ge2H<{iDJ4lNx$Y=VpS+DKdJI`)E8BY zNi7vvwhgLGy#d8=wuP->H*FUBTzS+G62z$#a%BV0q5e-ZY9BwUA8M<>wNgMGCy@OU z?$z20NU@H|e5OfPj$U4u1B+9tvG!E)EXXMLU>s{&rd3hYMg^Rv!R1vnioe0%H;#K~ zeQ)K^ryxt=KEI7K``Qu23>_GQCZm=e_V3$S$Wf3W;sP~Kfa$raU}tlJ|p)rY66}uerz`qh0Jb0 z#g_QGCpFFW0_P|>Z|%h`LdXPEE;o)U&E(nWg-Zz?kq8nv6r795aVKkB0=nm41_P0ya?^cecWf3c+k^NkyBq=cF}m$1v-A!YaVa_zPBhWrE)L01 zdW1Q|XJzLt18M0p+^B)>MABF?n}TacRu^y}qe|7R81H)w0adpK_?oJ zEYO|1Dyn=Oj*()NEsy5FO2R11Ecf{U+;PpRg-7VS1RX=Ch5HO+q+*Vr%+U<(QGO!y z0y=F*4IQ$&(=VheQ1V8!&Wm({hJ3~8PU9-8lVhztKlR05KF4JOE^qT}>tHZ4ux@2D z*=?JJ&b??%q(vz477##=C+QuKayhXspLI1pt*LXFS?cQ;r#0|#K~)0DMZ}dvJ}yOE zUCo1oL5c@m(WJ~MF5|8RN9s)`)uh_fX$q;G2-1SxMFz_OP)F{1bFvdN{#2qy{1pz3 zzbE}HPxp}?OQpYTB*9wR`#!3bE|uIQGt9iA!UQ+QyrMB-%?jkE$LwpZ?DOt0Q!?`L!)? zpxbEu?RL#V&_Oo9Y%xg(;ZE9{OJeLt9DqahcmIg()%SQ6EV?XFzQRF-74=Z~QlL-i z&EfV!4i?Xm`edP(aI$qbo-6J6UN*mR{1WHcbc*T=0n22b+f3D{`j%R9PGVj%TD0QV z6Z{}fqJr6&Ftayzz6aGjhY6=sI8J*|WmaZit~yus#i3jI5>ff*ERbQ%pj&CH*=?Ey zSwf8oJ|?zG+za7FD*lEk0n`AE1RhBYI!?&Xd(6`6J2+&z4mcaMLvRKmjMPS5&BpAW zr~zbo^z)&(JUWr7uTC5NjIGCk6EP?x6DYeavRtc*yba#CzS#5YR|39`=i2~aI9hA* zt$VZZL;Og3E6F>AVofov`gHKeJ@M^A#`rM~bHtnkvycMYiM4#9J4PaaK1^_^6FqiE zV(LgT$%_KdKUL%zgHUmG6b)wgD$_bKY_+reV29O+t7$6EK;mGW9q23;2MczkiKhEZ z+$}D66PPXtFgud0d5Jz9r7nV+a>b~(2*0;!zcT*Tnz8yKE?@Ez$o;wNV}LWfZnpL8 ze$4{#WrBoL%g>t=X8?L51ZQ%BTYd180KhjVubc(Q+TxjT(E@tj`_H|H84raMpW_W# zI_n)c1@$fHbcuF=Rln9xk^!rq_9a$rFTu?+sVe=}dfob>?^eZ*0(&XEO$n;wSvHce&!ZQ5nc|A1 zu!{loqECUUWFATEn=b((ZVoyKf%G^yd=YeBR0Z7nGgV*I<C(Whp zYeUcWR@Uy_EX;H$EDwX~N!*74>2>E!ESih_hK|(Qc{H;@TQ4U!_~yac9AfB%N(tAf za$+J;hLk*0!bOgfZ7%1oZs9~&P$6|pVl9yoIzWQV_=fnHMf|;9-;JhYuC+&z1Fl$3 zIZsMZXinaF7=4jIp2;`bk(N2)P};95?<&q&eZl8ezXW~Hz7&QoL$2G4TispukY+*n zZK7GVpEA@^tpB79#UPp*+nOX!={a=E)pBpCJwVOdtIm4RJ}nBfPZ%!UlN55^Irq7r z+`EIDW}(vrJ=Djaw!36Z%~4ACxB?lk=(WVt1qW%0gW@>rR>fbn7p^b-Sh*7HJ;THs z5DaZU*a_C5S;#qrNJWAhZo=1h(I7+C@h2|q=!!$pY#yf9t8uev9V}X_v=bWn2eE4I z2uNMknB*h?x}YW!h6wYf6Ox8EVN@g#)4_?r^7NUUJ%JPPFyCB=_knPSb>t0jZYaPY z!A8Nv6HEYIReV*MminU0EpQ3IHk)LFhGFSBcZWT#SwN%;!_9S^gzV>VxAm!WO#F5> znze2+V#|6k8%4K+`D!d`b86QuwTInE+Ebfu98c~@A?y|Ea}|_KMUa2I0!R#(ZEu>C z0uARGIZMttM;r{gC)Yj>B1|;S#VYc;|OU2?le-erlmT}i~WcFHjB z5!@+`arLzq0Otd}+@KV3qTsw=j^^C$Lm2JlS>8PN@*S%h8c`_OV^o4xh(Cp1iOMMG zm2Kg|tt$C8nX39iQRlD>jQJ$%MtQC^F?1P1H{Muw%Vq&qw(`_PtFSU@_-0qEG8i4n zwE(>rgmSG$sy0-8(;8x4Sz%|98kWMA)H;vJfLH#CRYg(c>1dxe9Nil45s*@&=nLUA zdO`G0cS%Uy{UsU+#h(>n;DcV7SB@rI3`2g(o;WMD24$&$e3rR4z!(O%ux0F~%|hX2 zHd-W5sQ<-pKAdOAqKmr z?-1+TICNniOog;Iyy*OWH@r~t+MV?4YDh_&SAjMnI{Ir~x9>ioJ_$9*$-Gsnq=Gj= zXxV4+S68^?`!k_;-C0U+K}}^`3Pl;D7ij`c?466c3bI`Q*-o051fv?WOIOOX8hkNF zvXQ>{a)8UY)cblvsw45Ns<(pFC(mYm;VFBe=M~+UKXCcNOAeY*RCA^oU|BP02uVZbhPKT-zlo-J+P?lW6XR=1U#}$%|t$?!Sws zaRqhqaVa2=q}YPiD4Li)i*?4?U9bF)0l}dJk4hM7KdpY;%%8i~eAE}ePpe*(zjW2f zHf$Imx4Ap)?#)6TTwAKla;m_^p@vakC;>MIVlwy5gG1-OSS^K%9Yww_IzI4w=rNR1 zZZX~qiYaw{F{Of#)zJuDp$<+rguf*J$bj{lm$On{z4ghLUO${-2>O%|y)8#sXNrCA zOQ~q5LOo7@*9*?4RYmL;X0X1<)%huty?z3J>Pi_D4d|P1Kf7nM;4O#2D@p+Zfbdwd zH6)UkQz`^d>1rfPD~<2wh$^l4n^J$&!%jzOHyc1g?<)N*5!yRlJ2xloq$z<0j##5T z6J`wz$gRBzVFV-GEz@H3fQ-zYS-k|^)v8CheH5&Y=n3Y{n0@-cPzhTO7 z{H%M)9@j7k52AZ$@F(F;gaDjIh5CX~ma_zmRVUa`V4&OTCa~K#3uNb6 z`c-lfk&U@D6TY~!2_XaBM~au_4TUn!Hw<te`5^ zDo;{i*n=7dQFtVh+YsMDxG;vmPC<-aUKy!`FT>0y+QVuo zsR>n!@sP&$XgwwMSA5dvtFSvr=aCRV^`pDFiwWQDNB0B5Z;rB2PRbaIx=5H+4l;fq-B)y9=Yk5`X{hYS6hH_*7lnxp*IiZbGAOc0Zr|67oKGNynrVIJ|qE{@6a z2VV!tu&OYt{;u_flB`<^luGB(wzQ#Qo9k(JZx~>*uY_oZWS8yjwok*o3ByDmb>vw<5XSK zR}Ve+UJLapk@(lW`YE&*Z(JhdYMNg+g1-ei9);v5>e?U*kLOevb&iCcnB@B*cAGqh zpX@Bj3~paXf5?xF(4JDlurrT)Ri$T@>8LMa{j!#b&Mi;7fy5xVg?(Z-Z5Hw_K-xLn zPTv;B8LXL}S+Oy5v5r*^Yh3m~-4gyfup%%#5(s zM1>_2PE|tBN$Y?*Ed0L{|B*=~iG+Ce;G>u#(fB*hp(vn&ha*^YASyt_$atLOq1bvd zmpt7BkDh-2J@2&yYgacoR+@wQLPplI1Tfp0Z3BeiXtOP0_iGlwI>}tlZtpStd%{3Z zSnZM)V44>OPpw4uNP+gepXaUi$qvi?#LDk#dqS?I@O!{qQGlDI{kc&&#ZE2wL}OCW z#VK|i;ITSJ_}uanol6XqgT2Ld5kkUkXwVXGTva)>+3eI8gDiIm1fOFC3_b?OE$kb+ zX~QttPbBNyC(^&`b_WP8Vjd%s+zI+>?Ve&}arFTo6Mqjn1=>k?cD1S?0&wAt(@3re zKX-c1K`8nStjvCsa1I}eORkaSzym{MAE=spbN&vPb1AtB)?*Qs!Yc1R36deb4nDuA z48YB%qP_@xY9RDhX5VmP;H$Vl>_*K3+}S++LJXGpU{JP@gP~S)4?D;0lhx{_YnB&% zrFVI$pl_9g<%hVSkvKl2qYOT4n zeQN-C@&a1dW^MUP!k~JTtex^>jTvDaO`3*C>?3keKx4EGP`2lfGlQ7Q@J+CT$#E@KWu z`f*>6P{);{3#-jheWB#mx&%pEoO6SX0q^Jn7ENn*KqpgCr5Q{2Kpv-O~`th$LTR;n$>E-c>DTY`eqadh*2Y=-cpG-yH1+{E7+vg( zOiCyaPY=BirStS!XB1I54-}oIQFR{m4c@rEP|7E34y~PLg$rq8Pq>OijIqda+idO%#Q8=gbd2Vv=oM){dbK8hn*RK?%lw52(4Bg)-f zI%}W$JTQ#qRh7Sr(^Frp$+VV9tL;v^fySV?^-W`sXcjIHlUS0#!uFofPlnh59p}Kx zNQ?Hr?n&GPoO!-k5J_lVFTJ-sFDKKMjfB2|rW{B)1YzG;ay=ar$|r+RcwI&}DO3Z5 z{7Qr$F}Tq**ngp4o!(?~aY=o2Ef3R(gBToIt_Fw{YMn?p?;gM;ucdOT@{{MIzR+Ve zON87r&9}kB5V)nyVmEFU?v7(I!HY%yI2ng0>75pHoQBle&@!rz5vS7{J{{kBnJ8V# zR0RikrM|1}sE%A7r!W%p1%#7g0@cq)SUY+>DUf=OJ3ucHrw`6(H1{->zY|{}LmhC0 ztSsTGyl{OnrYl#1GWj0=#@1%tP+;(S%0{qW&B6`=cCZwuL@%))b5;%Vs76z45>MjP z$QuVaqef4P>`3g>ol2-+=*topM3oPc^^aGq_)G7TuemySG>KUItE^AHq$`9vuBsGP ze9!u##w}P0Y^u(%ZFGZ&_eMT>q3sUMg34=Bd_#z#2U4&;>ruEA36l=eu6OlPqJyn2 zri{nxQY|$PmWbOpU!C>S85Qnblx-wR9HnfiYlD9a5!H(aiV$ZmELN#1psFxBN}Ebj zvK+O}NX*JwZ}6`PnkBbmyAlZ|-)}nK zw)qWWPh5HH*eujsX4Izv6z;eblfx0+$mbK7WRP5E6T*ldu^k>;YebT?#;+@_7idu{ z;+x@kl0rIo<09Mh9IEfF6;7}B15)&;`&X!FMlBMJ^^(++B9NkW?D_d6OIcpv+Q)2k zG*Pj-alYbw)E87wJ56nA)(s?vt}Sd5yJ^Fa7h}^7A$19l(mt_+c_vl3J4pCDd^bja z^ej%RW!9n(qEOBnO@n#PJMCR?4# zQBeAb%CB8toXH}V$gV0=Y#ZOuw8@Iw-I@iXPV!Cc6~tP2k=3D?gh;TJYM#Yd><=)k zUr%77`1d_xWpDa3=lPNhtDaO_EUJ1ZXl(9Dii8e^aNx3oT}1NaPn;S3AREtUcaoAs zU)3)0^w+5`6i%WtB&ANT4{`-!ga)lz4<5_VwfPt z?q(7VAJN^Gy9B zM}#rG&BS0{8DtyGUVUMwd!a-GZExNUREEf(yq|0U%|bOwLRY*;5~k|lqyg=OLQRvN z&_Vf__tmm(sRyPD*CS8G_a`waYVV42ECT5Z*wjjeeiVv07xcc*Pl{IzPb4?-h@YKn z=PdT=jR#rswX#lF{k`jp#ve!M=5M|JJlpm+{8e0gyHT@nL`V@7>jI-JG&#%OiFag}+hGva09 z`wJ(5>qL!9A#JNwP<+bkrLZX^^_ zk^xGJSsPy6;3QXmnmV&r^zj$GZN-b1HRA>kgWmQwgx$MY(1R8KMhHg;m*#&LlVm&~ z-A8KbhN82s7OcJ*+ymbYmIG(De^Z~ zg-(9UC%;X7VWR6%BHeyChqhr23>B8mZqlq>k0oTrQBO8q=)8XTDIQROl=aSRZhhf2)Q}* zP;KdU271ydevpU@;m6K9n180?NbLBs@T+AL4PT+gNlR2f^f9R__^zkf;PV3B+?~}& zzv978Ds}tm=qjGvh>QQNXnJUW!SM_I@(eRCL|=0$7K2yBb$}FKPkaPRZ=wTkVKVBA zMZcsaX#5#VV=yv=mTnzu-YnF$N4*KfvJR2~+k6&&Ob0L`V36#FoVdj9BIb{DF?(lV(49H2KXHUGCRYl*1vr%8jTBwBbFvX7#^=w}e0TA6) zg-Vt2*B2kU2qob7;WXMVU2OUvE|m3Y*3PET3ZWipc$o+@7A8D)wMHq6nPH_jT${<3 z&YMX%=M|dPT@pR8)=NJ2AtBFH2qf(J^;Z&kwK7FPv%BP&BnQ?W{rZ9_R{r;L$xgXQ8rnUj*^I?-NSV`StUax!xl-%05`dC&W(axh6gwDsqy zz6g{hFM;6CS`tH+!F8i;Ww&b6`oMcyc%5JO5s%nF-FJkg}DuYel=CQf7{b}2M^#WPD zX6@sPD8bZq6N*|NzCM0^&_2C@j}aE$k>*q%Lp?g0t=`GL`76}1ynW!>LyAwd20!~# zXBy|O8X}d)H)=UPjhDjhL;z%$iq12v4_DdjfgqA`Nm)uArOFhsj4R`*nZG4Fi%>ht z@>K;!)mf-7G;}>nq{({oYMb9c@=R-Q4{8?54qcU@ov07guw$hH0Ox4>M9SDB32%8r zhaqkFipZiwfRz_X)_Z+(vHTXek7)4~D9$3>y0dtR7_J~_bU>(VIy|Sbi3-<=omQt; zcKclUl#cGv^4ziMD2``Q91y*9D{iQ}yoQT}M=f|z1(mYPzT|ozH{wojqxdEx=8-C6 zf00sS|Lq@9(4WND|NQG1ht_2JGVXTo^auT=yP&^4{_mITPh@=|>UKj3M*p=8mRI8)4W9&P}oFF{mQh#O*sS*tIwb?HlRz4;^?atxq<-u|(% zGz-P;XxdK*PUPT)Gfax%rcg=7;l({X`q&^>!&sU@B1tBXZx(z`N3m4j9MAeo$>p9U z&sCuzNyC-jXkm+-#A$&Nr68}}U^vRS!Fl;C1SnFQlYd(K)k|fp|HgHOpsrde2&Kgj zrqcGd;cWX0Y|m&Geh7vgN06HI2-fbPDNeC_0RWW6qcZOqz((h@{x) zi+=HuzecmZL+I)E~ZPud1ozAD0`C9x#dJB&w=q+S4|8ux=p1s{o4<7kKZUnmru)PdcI040e5B#UjA>h0`iIRLIa^ey$H0I2nY887cwK+Fk3 zNS~=tA!kw4K8Pmu;{nze%q^5*^Hv%cx^Nk7Q1xk}7pyN#Sg{g}RGwwq<%W+RudnrK z7Ba-+`z;QC&q6)n7& zYiY|1k}-QNu_xk*c0pDJC`kzmH%M&`V~{IpG4~7!9z2;)AZUy8xfm}`qf2s$tH z;w}nd52H+O1%3OXs$8lxVf95A3txh;=bC&2lL7D z+xG^tZ7jLnwOMGwZ8k&?+>wANIZw^-%zZoy_bI?3r!D>5(jwEV(PUj-1=>f6gT%Y% z3R63q{uy<9lkbBe^`YvyV+e{FC)D2iPe&vzPlC<_vj6D}gNNiaW0G+>&PQoBOtaZE zf|h`QNiqYfV{4O1s4uqMnwChpt^bl#nn9vd~Kk1k8Bs`Nrxg}hx}!V0IaO}u1~@s=pL-tx`e-ANc|g}=l!(r zIbuKHLbyfKyre!=GK+4VDdP}}Vh^}oM6+o&uWZVw^4j%V`r^z%~G7qm<5+ODY4$K#FN@wtpibE*Vjo`Xvr5<;B%w2s!sr#S=lRtTF40eXv zt!+8Gb+Zt?+s~kg!wy41-)^auop3*F&T_)qqQL-MS zX^#}r5%kKL+7MSuENO;gY!nB+*(-u4?<3->1N(W?;vP9Em7q-8xA3_MWm_-!`0nM$ z&T;#sAgUZIs0>G;^pPry*g_%+{z%fGUD?>b)(h4bHZnbBkonQ{+4eP%Y;hs&=FP%} zRx*PqQIeX)fG9ea&}k50EO8SdhqskV-J+>C9qR81=wE&b zF+Aek?%sNM2r06JXxzo#X?8-M1Gf1|yn!C44_)hh>x&p&wGw4+3lnYo-axhP(%T)H zg&fWpZWryOsr=-JB%FyiS#zzegy>4}_#Tp?=6Vo*zo;X6DCD1u@0P08>lAe`EgVIn6N z;x44ngpjRqWwj(J^uRNe_t4GEZ;ih}us_J4?4AeFEQnmT3I+wT(F#yF)WbjL{DA`l zR)sG}jCS=Z6V+d?zKCF*N`%Jab7=e5FtO?4*&UmOh$G_nhGDxuL@b<;6U8jugBA9Bw>G; z;78voS6<)n1?vk9w`L{ic*3OGrZ$*VTTHu8v%qm7lHpR=0ltoX;wVQ*yOv>+0JQ-mzs|ubf?_OWv*cz6gCvUU8xi?rCwAR`g*1cJ%dYeSz z!sM7NK*~c37G%)Qow-scv)1OO^gc~VNrD$cq7d@Z1zM;JX3bdw=nf`nHeU9_6i}x_ zQ3jgahw0Anhh7lr?`nH=#ef|m)qHlDfk$`2-{9Dtc}TKA$_cAYb^e-)lT=>>y5%l` z-!m?P0m*QvD^E=*pG8yDua6_F4CnKaq=I~Uaah@HlRuN+ zqV-?0&QO!LDTUg`Q*PVdAXT>D*1B0B*-ONi>5HVREi4ZK7j_9)*zX}r7a_O1`_Zk3 z5+J=8_s4(z&1Jtkz8B035=c(%=0z_(Oy$uy$h(^%<^?^S72v}F2%U3+=dB!{0Ip>W z9w$@bo|)eA#U5q{C%uSf9kt#OL?@T>pgFpM!o#ex*}C#%)E7vyxFs;!`qUdt3{=%O ziQTJNkb6BLB8{N%9ZIy3(Q;&j&j3vly>oqO7asF?gco?xc|xEq zq@;@><(v|AIKQIm;(j(2KkRuN4O|5T+)7r1WTiM5C;{?MSQbN`!Tg!_ zn?0ym5WkEt(TB%bD!hke50f@ODQ0GQv&$p8?xGWQeTdDN3crRG&r?;`MO{EaD&Q{2 zR33|7Dh#h$iX>=w-eom67zXQS|l?HZmXcO>k?&ol8pIzl?i7@V{{7SE;jm`0aN zRx0XLFCM0FUN}+>Wb7=yAA8J+lXgypR?(xmrgkGN~muVW6*`OL%!M(Ls5DyUJHW9 zs`?D8PfvYOna^&SY|PtiZv_lLhRf}39=mt5uqkzaZ(5&FQxCsR-2s6- z2vp>17gKI09$dG`AX6tzLDOY)a^mz@mE+^*zAMrp!t!JCCBkm2bUVb^y!iOy$dMB` zIvxtw;Fzoo(8|+OUy#aTmq2tIvu#*0^gY{7u}3uwmdHowM#Ft!oE_*Ia%mQTJ?)FJtpoF~zd;s5Q$gQ7|H0PcWX!tr#m$~cRw6Og|<`zo<)6Bt6y zK_9BN3Q(1nYALE#cZTPz_qz3k*Fh^UcMM!d4t#}klDNi-}1 zivf{#<3^`|;czwvJSdSz={*77F30l}a}Nm#K#OqGY%=QuGja=Nd=F8#gW|_{bc@o< zBgB*Wky_Iv0U-xF{p?oqL))WUxUk$qFfiS4=?wb0%D#||=d`|1m2FZY32U%uhEqfP zmUgGzxLH2{XnQTGU5f%M|2sQt;R%6BBX`G946aJBT=m?!={(gJ)4K2_px*KV7>*2}+ulC*gk~XC zT@K;{)Wc`zHl|4S+y$GTV+S3~TkX*(CB%-?Xh{tEaz_Fc9`y9_=&sQNBFU`R^R-j^TwMpf=4W;n66) zAH+SP)j64trg%P9pJRRdz}m}6R@=KUTn${G1@ErDAYy&)7Jiz7(MaNlc)Z&d3Qb~! z^plsjVz$-K8ZKml6?rQWyrUbqSSnkMD!g}nVaNiO$kVONvu%GvPsJ@@H)rY!QeQ_GLeb~E%{jbK<>W;61wlDl9Jy9vM!7OPVM^|wkfglbRYE~7n zbsljpeG8SX6MfFmDq~bU308rSvlY4TF0)*W?#IdWn;d(EP{}p_E(>2TQ&~9^cB7MM zlu{zHs({>Z=IV<|)}RcQf5H+O_6)=u?m4?*v%vg0LB>dcuHrdQ+cFB+M>)pwe65AR zlkoRn|MTDe=l^igLdBD&S1#St6jD+b27kufAydQfkaKf)qEv-d-I4vQGG}^~SF10? z+zOS*zD=gsHunN8ek(k_zSgHyXg3q< zsegd@DQ8R7yZ`Tkw#o3R4UGRp-$rqXQix{;81i~mLASx&)E9lS%q0Tr2UBS%Fbw@^ zJHUq0EI=McSVJgiDw&F*lTw7OE~3{~pz#E+lb45e=&b7k3uS_H;t_WYvcF9TTE5S{g2 z1kRVt*%8SXMiV?iA0)p*vE-v0RbBieKUQH2(F@u_^v2D^*5-3lU#Q9Smx1jkE{`G0 z;QF-fWc`|jNCGZNzD0u^VQO5KK;EgYwK`VnkZshqg?w|(IkUxMDlSS^a;`|~xh=0{ zui>BwF^*v;=WhzyAVbjj&pa1c4`0Vg9vvf=dj&12!R|okL3x=uW96;{Zev}(@mw}% zQ8FD!!K=4vzxn~L()839h_cQlLho7Ezz}4}+rk#In>GuU?ZFjMZsD1f8chdL*d9f* zl(QX?Wm5mN4|HOc~MvjETa{aa7P63|zP>5&I zv8mK+-()>bYi+!MBPx4HwAu+#ZBz*PChg+nYjE;HFL#74XOJLwU3cGD*m={|X?l~n zWGRUlpLVNir<=`MeNmgweHnD~Hu(fU=lU4v49Q#AZFbXUAz93a_hd_nf*3+$J?@9+ zt=-ep5Hrc@L!mYbw2||QLwFrStw>kOb&q0K2C@K7r+xZI>B=92G(Hm&>psQ=fWzZ+ z?Oo+N6yqU&bgw_j?z92nZpm$*D}=^Xh2MIwT3@*7Qk6)_?asFC`vsVEseiNqtY5S6 z;$=xF9PZKJDq0vb^{BWL5s!^Kp$ii!9Kj;rR-;Hc-o&J-r$p_>=3eTi*z)Y!Ld6^* zB)MAwSMFT0lOo9!C90ENCiB$iO9%31Y%VnZDS7V3$Ldg1}jgx;ll7# zbt~ANnuV~_zQFylPyT7MYxpoc{>&|5Pj41F z$rn}Y=}9bwPk&GbNe!p2<-+l&lo*EzsFAcIak5`b5+UHnFxJ`obii$1-g%Z?pDX8!C#>+R}coyEhASyU}ET z$>VxvTxU|4xZ9Tu8=rvwNC60=VbS*D0L$B}f#Vp5SnDF542dMxLSh=MwTjVv7oEBdO3aUc`a&574dI(?Lc3eD5d9v1 z^$2gPcwF;Ht`J6j{d@H>Ow5xR>fJ&C^oi&koRDqFH@xTSm%KeLxvNteA5Zrh9~5 z3gi7Q{4gsyLi2&3G%+s4;65mw&&0E4eu7T7PsJJ}D<1N%JPWn;;6+t{RiA+RVlAJ> zGQj0+DonLO!SJ+&ePB0j7QS|)!5wG7lWfvUySMHvD8XDL zV~Il1#>x@c9w5=g>p%V4gNvI!F z6;xHfWscP>9ek{ivA&Aj|u@ zTrSS1c=R==408m`)Cm+oDt|}@wo4@+PQ#V8jKL5s5;uDwg5 zU?-*SPc)fZv|~Z!2$Q-#U7_L$NG`L52IzvJ^0-B(q=*e(+Y`q#y1_G$V%m_4Ri)5+ z?_Xa)v4SPiXS;K4P%wyXuqW(h&BE1D93H=+)QDQhB}PPfUnQHnOWD|rD=YOi>kzmO zQis-eAmasZ7&5AWA_2*fQ`^q;SVZ;Mx!2=5s=B=;2+Bw6>*vR*To2RfI7YS98qt^w zVP((L)jT|fN^%pr$O^MkU&#G9*zo5sjG@V(`IH@Hy_$x>#^OsW@LacYb1s>&_B z_uB7PUohz^l}NCs&!}x?gFuDVw42ll5ifg#-~PgmN$!68%m3#8{Y!q~FaP*2tyb`l zmr`-=x1iOc<#n1$3}9F4PyO?ktDBa)@C#pYk09y&^Or7%AvW+Yc_aNyzVak*mS}d8IC6eUH;&%aDPdwrMvmHuz%b<|Hj(rnus%EJPW3i?mvIYU!+S> zq8DHOAkTaLqMHt1x<6aLzW9$9|L2S8AQ`^Ew({c3^cT1S;R|MleK&gk{OzPY;k zc=;i||JQeu-QT10+v_i^wiF%2XO9QvpT9snJDLxvy!6X&|9I)f$`0{=;F<5G_b1#sB+XeZ%70O(nv$`yyY7mu$dzHhuYJ`tmN8TA==Q3orWH zKmOnR>3{t{Y#{60zDE?UJOj|qBrI?I;))Z 2: + # break + words_weight = text_rank() + tmp = sorted(words_weight.items(), key=lambda x: x[1], reverse=True) + with open("method3_dict.txt", 'w', encoding="UTF-8") as f: + for i in range(topN): + f.write(tmp[i][0] + ' ' + str(tmp[i][1]) + '\n') + print(tmp[i]) + # print(words_weight) + + +if __name__ == '__main__': + start() diff --git a/test_textrank_zh.py b/test_textrank_zh.py new file mode 100644 index 0000000..a026cfa --- /dev/null +++ b/test_textrank_zh.py @@ -0,0 +1,18 @@ +from jieba.analyse import textrank + +with open(r"D:\小工具程序\pdf2md\output_directory\good_i.mmd", "r", encoding="utf8") as f: + lines = [] + for i in f.readlines(): + if i.strip(): + lines.append(i.strip()) + else: + lines.append(" ") + +print("".join(lines)) + +sentences_list: list = lines +all_article = "".join(sentences_list) # 将所有的文本整合为一个大文本 +keywords = textrank(all_article, topK=10, withWeight=True) +print('Text rank 结果展示：') +for word, weight in keywords: + print(word, ": ", str(weight)) diff --git a/utils.py b/utils.py new file mode 100644 index 0000000..a7afa00 --- /dev/null +++ b/utils.py @@ -0,0 +1,157 @@ +import undetected_chromedriver as uc +import time +import random +import json +import matplotlib.pyplot as plt # 数据可视化 +import jieba # 词语切割 +import wordcloud # 分词 +from wordcloud import WordCloud, ImageColorGenerator, STOPWORDS # 词云，颜色生成器，停止词 +import numpy as np # 科学计算 +from PIL import Image # 处理图片 +from bs4 import BeautifulSoup +from lxml import etree + + +# def get_current_page_result(driver): +# """ 采集一页里的所有item """ +# result_area = driver.find_element(by="id", value="ModuleSearchResult") +# current_page_results = result_area.find_elements(by="xpath", value='//tbody/tr') +# +# names = [r.find_element(by="xpath", value='td[@class="name"]') for r in current_page_results] +# links = [r.find_element(by="xpath", value='td[@class="name"]/a').get_attribute("href") for r in current_page_results] +# +# items = get_items(driver, links) +# return items + + +def get_items(driver, links): + items = [] + for i, l in enumerate(links): + item = get_item(driver, l) + items.append(item) + return items + + +def get_item(driver, link): + item = {} + driver.get(link) # 获取新的论文链接 + time.sleep(3 + 3 * random.random()) # 等等加载完成 + + # 标题 + h1 = driver.find_element(by="xpath", value="//h1") + item["name"] = h1.text + + # 作者 + authors_area = driver.find_element(by="id", value="authorpart") + authors = [a.text for a in authors_area.find_elements(by="xpath", value="span/a")] # .get_attribute("innerHTML") + item["authors"] = authors + + # 单位 + affiliations_area = driver.find_elements(by="xpath", value='//a[@class="author"]') + affiliations = [affiliation.text for affiliation in affiliations_area] + item["affiliations"] = affiliations + + # 摘要 + # 如果有更多，先点更多 + try: + more_bn = driver.find_element(by="id", value="ChDivSummaryMore") + more_bn.click() + time.sleep(1 + 1 * random.random()) # 等等加载完成 + except: + more_bn = None + + abstract_area = driver.find_element(by="id", value="ChDivSummary") + abstract = abstract_area.text + item["abstract"] = abstract + + return item + + +def get_links(driver): + result_area = driver.find_element(by="id", value="ModuleSearchResult") + current_page_results = result_area.find_elements(by="xpath", value='//tbody/tr') + + # names = [r.find_element(by="xpath", value='td[@class="name"]') for r in current_page_results] + links = [r.find_element(by="xpath", value='td[@class="name"]/a').get_attribute("href") for r in current_page_results] # 总报错，不知识原因 + return links + # [name_element.find_element(by="xpath", value="a").get_attribute("href") for name_element in names] + # [name_element.find_element(by="xpath", value="a").text for name_element in names] + + +def get_links_etree(driver): + dom = etree.HTML(driver.page_source) + links = dom.xpath('//table[@class="result-table-list"]//td[@class="name"]/a/@href') + return links + + +def get_news(total_num, keyword): + driver = uc.Chrome() + driver.get('https://www.cnki.net/') + time.sleep(3 + 2 * random.random()) # 等等加载完成 + # 搜索 + input_button = driver.find_element(by="id", value="txt_SearchText") + input_button.send_keys(keyword) + time.sleep(1 + 1 * random.random()) # 等等加载完成 + + search_bn = driver.find_element(by="xpath", value='//input[@class="search-btn"]') + search_bn.click() + time.sleep(5 + 3 * random.random()) # 等等加载完成 + + # 获取相应的链接 + links = [] + stop_flag = False + + while not stop_flag: + link_current_page = get_links_etree(driver) + links.extend(link_current_page) + + if len(links) < total_num: + # 下一页 + try: + next_page_btn = driver.find_element(by="xpath", value='//a[contains(text(), "下一页")]') + next_page_btn.click() + time.sleep(2 + 2 * random.random()) # 等等加载完成 + # driver.refresh() + # time.sleep(2 + 2 * random.random()) # 等等加载完成 + except Exception as e: + print("没有下一页，返回当前的采集的所有结果", e) + stop_flag = True + total_num = len(links) + else: + # 超过了需要的连接数就停止 + stop_flag = True + + links = links[:total_num] + + results = get_items(driver, links) + + with open("result.json", "w", encoding="utf8") as f: + f.write(json.dumps(results)) + + driver.close() + return results + + +def get_clouds(word_list): + text = "，".join(word_list) + wordlist = jieba.lcut(text) # 切割词语 + space_list = ' '.join(wordlist) # 空格链接词语 + # backgroud = np.array(Image.open('test1.jpg')) + + wc = WordCloud(width=400, height=300, + background_color='white', + mode='RGB', + # mask=backgroud, # 添加蒙版，生成指定形状的词云，并且词云图的颜色可从蒙版里提取 + max_words=200, + stopwords=STOPWORDS.update(('老年人', "的", "中", 'in', 'of', 'for')), # 内置的屏蔽词,并添加自己设置的词语 + font_path='C:\Windows\Fonts\STZHONGS.ttf', + max_font_size=100, + relative_scaling=0.6, # 设置字体大小与词频的关联程度为0.4 + random_state=50, + scale=2 + ).generate(space_list) + + # image_color = ImageColorGenerator(backgroud) # 设置生成词云的颜色，如去掉这两行则字体为默认颜色 + # wc.recolor(color_func=image_color) + + return wc.to_array() diff --git a/小实验/t.json b/小实验/t.json new file mode 100644 index 0000000..8df455a --- /dev/null +++ b/小实验/t.json @@ -0,0 +1,168 @@ +[ + { + "domain": ".chatgpt.com", + "expirationDate": 1717557553.524072, + "hostOnly": false, + "httpOnly": false, + "name": "_puid", + "path": "/", + "sameSite": "lax", + "secure": true, + "session": false, + "storeId": null, + "value": "user-41eRLUDolErDqfpxjCML946x:1716952754-s%2FM4L0gOaiaBIHQcI7Of90034e%2BaBMYJQypD9vaHy%2BY%3D" + }, + { + "domain": ".chatgpt.com", + "expirationDate": 1724728753.008259, + "hostOnly": false, + "httpOnly": true, + "name": "__Secure-next-auth.session-token", + "path": "/", + "sameSite": "lax", + "secure": true, + "session": false, + "storeId": null, + "value": "eyJhbGciOiJkaXIiLCJlbmMiOiJBMjU2R0NNIn0.._O6IVC5P8jV1mAp-.05WwBPgBtboU8kRAEVTAevQVqdpM_B-736aGH-vF6H6eF74nE6PsDkcWCq9-n6h_5Q71-es1K4Axcl4ZQP3zfzDXgyQbo7j1K4cEtM3txWojH6g5YTtZslO2jvl49CEjvmAZGwb5OdidtTK-_h4mqi4DZA2fPWQG-LdC5I47KJuTRZC1_dMcS-Wwe6BjLWFJaIuJ7b1Zyk0i4Dvuy2YhssaVKrYBMJR67IQdEYOnAI7Rn1i8hcDuGZ9U784Ewd0xWkjwVsxsmwZ9b5dT5YFE_6OyP5UqYn61cXP5xHf-FkAjx1F_IPwdRHOZjiCCEFIvTCkW-6ObCc9yja21WxqXBCDVEUyGHpCu7tI1hkooum-28ViOpm5grHCTQ63AQoI4JdPY6muIUs0JFjFSO6KmyU6mAAxn7V-nLOui70cX7nc0rI0bUNDHaWY_gh0cXjjAgxUHE8UCMBZTJFnZLDXDA0rc6bWSHQnd6SFjBSEYey1nVdCMaajOxIWWr51TzQpBuPwJUT1DGy--RRF_mgX1N_lRiu97yPY8sGekSr77RWoftvyDzqZfJa3GazKvTBsTYpi1q0JzNs9KD-ZAmpAmelTxT-Q-_wrPcGxVzNolJRQZm7SSYq5OjieaY-VAKA6CpP_ku738CdpbCFCNJ0u5iooDd9kqdz_V91872-YzJcNQixw_2Iku9dh0anHHX90BoBrPmaAJ21cG6wmA1I5y1TaIGiSlPcaKlKUTy5OJqdw7cUhaEKIBShc5KF1U9YMY7Z7oDahUptoMO2EYDgCCK4nM1InLnOuNDW1gYjinv58EcOP9WBmDLu33aqoZdFgtydmwFihBpg3ynbsOfH_53PiqxA-0nI1vdmeGR89q-sQcn4YhM2z-qwxCTUDMuIYCA6OYgi6YzitEJ9ZNgcuP9Svcux00yfnGki9wT51VmCFboS4K1Nrcx1pQQpjfLlwSiKs9SBxiKOLe1h7VKbvdKgSAzEkLgts6MZPsFmPfPvmLbu8-Iw7zT3WIeicx4kJhbkTCu54xrriqPlUkDpd5-dQKD_cunanNYvCPCsbXGNsCjmIbvvZ1wXeLRIP2KxdnzwftEd0KtmhLETYIkbLojqdnZN5wCPWhpCgd8_yFm2PWISPEbg7vf6yFi98TjJBWliwPG37Tqb0-NSq652J3XC9nZSwfhGkfDL2jYm4vY3pnAoFzifdUNMwNxyRHI05uFvWyEG-K_ry3z4rYEx4woBhj_wSbsqzpY3YMUxJCcSRaj8XqIf9_Xs4HkSr2XNO_0e_ObeQUCzDdS-zQ5G5NTCTVvsRv7GDHIMpH2sokITtObhYzweIjmqtqxxAauY_SxWm2GFoXDocOFTYdCvOuTvYoijc5zTAbT-i1f6ZjeZSOqnJmjNkhsgSESQsNtNc6mUpbT5KE5TatI9pcpS-EPxLVYds2ROGlE2NJr9TawKJTswPBsuOxe5NVP6zUuukDxmwNgngOtL2D3GyUvGnEqi9bKKFLyIrSfC1zYCNpz5hKkIRJn3bvkWMQIzXY8-lkirYXcZKZkFdKR-6lm5AOe7lSh7tpqMb5OW5YwHHKXKr9-LmX1rpx7F-3t-rcbhDAIs4WCPBlKdJEfwkL74qRdQrKG_idgmFfCdmt-L7_UAWlNlIlcuJzjXtz1BZHzizJeqCI8mgtxtmpDNvQaKs-Bxo-WfYMbUK_AxMh2b-S13VWy5p_gfTjnLpeal1ASpjf1Fdu9mvXNtRoSJpgkjvAKVhwhkJLEusGjZ337v-6-sffEzS2yGPztf-KhxHvjObREEyo_eQKV3tRGpLIEOCusT99x-DnXKoUA-QMI3ZDaRa8pQhSL2UhHofGvQ4M2Ci3FPHP_OPuuy_0NqZpcwe4ZaNKWvyKsoJb3uKq4Mppd8Pfv2JRJZo6SfLb8AUxMQgNsbBRtBPBSe68oQ32ent7JWanNngcCFhtAJVCU-9ACkkgHI0ZRS4pcGi1WrHJkIXiXTHC4ZCeyl6w1R-9TXuxVZNCKuBUNUJ2Ny2OHi4JZYafoA85vWdSfPVVvzWN3kgL1yqWGYA7DZJhRqmG65ZKSSk3g-KXjoK2gByEm1r-eUYwbmctcTSdxoAV_-KO5BgxHF31HceX3X9DWjEm8SN8LBQtwzjc__vaQb5RsqjBAtXYsAVkX1PaDZnDMY0sc3yl8KvznAfN7cINkiwVSjFhBreX8lX8GXf27d1nqWcnby9ERj4XC_Om9HcK17sFLdgXYWY0BFB6XvoAlpWD7_lIl_-vb50zxTxMOjIVnfOEPQKjZayB-ZYedPeGVkkvJ5gXMhSXnhnGnvyHF2C17-LymmvjhHP8Mk4DKVH478AF39Yh95uPrD5VQlJ8WtKpqTEBFmyaqrKDr0TSL_Qsuv1gnJ_0-Y_qv00PMhTjfa4j3cVPUypVOsfqx7k-xKw8xChKd3l29d-drx7dwqPd5E_cW6K_EWy_D7t1vVrLXkrVjXnU1naNuAY5JRC86WrHVYS3_Uu6KTVivlqhoERr_cTS8LQSZTbiJ_9221tRIIWcvg3YIeFjOzs4MWd8uLlTBPLhHs2eNVsC-xENfejuHq1jSDBhq30QIhVh_G4lPiPikZAZDCs0xEkomsirA9a6SjDjO1YsE8SJr1KjNBz4i7BWloKMAix_S730YwZRXXk_6U1cdUc2JdG9sJYOH_Ebyw2rxUw5FJdL1THDTDPVv1DBPutN7G3WdrAV115BseLifUvp6NqirdWXdesKvcYvI4umhnm_SjqO3iUrHEgZSKdLPzpf1JvQta0YCcq4uKuKTnZ0ty1qYYwlRvrcqaz39vBcRpcZ5NmZLL5JkdnxKsm2LpzhowRhdAkQp0uuMQFI9a-daoCqMx4mP_h8oKCrSNMGw2Ob75t1DDX_-Vf-oB8mee2dguwyBOeyPue8691MCLLK6iUIC_cQCYWdcwlAcVPdlziul5e1SDjC7Q.-yX4k4OiubAKyq0VYZa5og" + }, + { + "domain": "chatgpt.com", + "expirationDate": 1716953115.64912, + "hostOnly": true, + "httpOnly": true, + "name": "__cflb", + "path": "/", + "sameSite": "no_restriction", + "secure": true, + "session": false, + "storeId": null, + "value": "0H28vzvP5FJafnkHxih2mSdkpUZExZgh8sAggEBQHn5" + }, + { + "domain": "chatgpt.com", + "expirationDate": 1746093367.413939, + "hostOnly": true, + "httpOnly": false, + "name": "oai-hlib", + "path": "/", + "sameSite": null, + "secure": false, + "session": false, + "storeId": null, + "value": "true" + }, + { + "domain": ".chatgpt.com", + "expirationDate": 1746667743.898085, + "hostOnly": false, + "httpOnly": true, + "name": "cf_clearance", + "path": "/", + "sameSite": "no_restriction", + "secure": true, + "session": false, + "storeId": null, + "value": "xvS4EyBgcyCGuzaDherbVJN.4PTFypiHdMr2cs1xyvk-1715131743-1.0.1.1-PpaIw0m3Wm7XXC7fE_5sbrGPKyCc156JF3NiRzk3lvb5zy9UWG7FwZUJPa3hEUg1eN9KsAPlf6.L0kw_gMKOfg" + }, + { + "domain": "chatgpt.com", + "expirationDate": 1716953679, + "hostOnly": true, + "httpOnly": false, + "name": "_dd_s", + "path": "/", + "sameSite": "strict", + "secure": false, + "session": false, + "storeId": null, + "value": "rum=0&expire=1716953652562" + }, + { + "domain": ".chatgpt.com", + "expirationDate": 1716954547.878219, + "hostOnly": false, + "httpOnly": true, + "name": "__cf_bm", + "path": "/", + "sameSite": "no_restriction", + "secure": true, + "session": false, + "storeId": null, + "value": "xvsQlddTcKTT1B0Yc_xsV.rNMDDSMUt1I_wLjaT0hr8-1716952748-1.0.1.1-VKOlrHicFOvqrsQ95b_H2v3hcnW2AdjSPlOe9.F0f24OqrihsZ4bP_BNWuioEawn5DGJq_a_LlxMoOdok3dQsg" + }, + { + "domain": ".chatgpt.com", + "hostOnly": false, + "httpOnly": true, + "name": "_cfuvid", + "path": "/", + "sameSite": "no_restriction", + "secure": true, + "session": true, + "storeId": null, + "value": "t9el16WmScbou.hCjoh_kzdOfHcElqM2hRATzaPQLLY-1716951316241-0.0.1.1-604800000" + }, + { + "domain": ".chatgpt.com", + "expirationDate": 1739533563, + "hostOnly": false, + "httpOnly": false, + "name": "intercom-device-id-dgkjq2bp", + "path": "/", + "sameSite": "lax", + "secure": false, + "session": false, + "storeId": null, + "value": "b8e6b16b-9cbe-490b-96b0-5c335577c4cd" + }, + { + "domain": "chatgpt.com", + "hostOnly": true, + "httpOnly": true, + "name": "__Host-next-auth.csrf-token", + "path": "/", + "sameSite": "lax", + "secure": true, + "session": true, + "storeId": null, + "value": "0064ee7bcc1a92b8ecb3ad93d256c766dcf15e29b40f4244c7174b63842c2191%7C6fea0827d7ed6463d7e893a77072bf348df0c2644b569ed302bf51a7b3502876" + }, + { + "domain": "chatgpt.com", + "hostOnly": true, + "httpOnly": true, + "name": "__Secure-next-auth.callback-url", + "path": "/", + "sameSite": "lax", + "secure": true, + "session": true, + "storeId": null, + "value": "https%3A%2F%2Fchatgpt.com" + }, + { + "domain": ".chatgpt.com", + "expirationDate": 1746093268.737159, + "hostOnly": false, + "httpOnly": false, + "name": "oai-did", + "path": "/", + "sameSite": null, + "secure": false, + "session": false, + "storeId": null, + "value": "04905aca-630e-4745-8b12-104824975054" + }, + { + "domain": ".chatgpt.com", + "expirationDate": 1748488752.141889, + "hostOnly": false, + "httpOnly": true, + "name": "oai-dm-tgt-c-240329", + "path": "/", + "sameSite": null, + "secure": false, + "session": false, + "storeId": null, + "value": "2024-04-02" + } +] \ No newline at end of file diff --git a/小实验/网页访问gpt-4.py b/小实验/网页访问gpt-4.py new file mode 100644 index 0000000..dbbbde8 --- /dev/null +++ b/小实验/网页访问gpt-4.py @@ -0,0 +1,129 @@ +import random + +import undetected_chromedriver as uc +import time +import json + + +def get_cookies(browser, log_url="https://chatgpt.com/"): + """ + 获取cookies保存至本地 + """ + browser.get(log_url) + input("回车以继续") + # adkinsjoanna26@gmail.com + # c1lO2NKEa2Hsl5 + + dictCookies = browser.get_cookies() # 获取list的cookies + jsonCookies = json.dumps(dictCookies) # 转换成字符串保存 + with open('damai_cookies.txt', 'w') as f: + f.write(jsonCookies) + print('cookies保存成功！') + + +def load_cookies(browser): + """ + 从本地读取cookies并刷新页面,成为已登录状态 + """ + with open('t.json', 'r', encoding='utf8') as f: + listCookies = json.loads(f.read()) + + # 往browser里添加cookies + for cookie in listCookies: + cookie_dict = { + 'domain': cookie.get('domain'), + 'name': cookie.get('name'), + 'value': cookie.get('value'), + "expires": '', + 'path': '/', + 'httpOnly': False, + 'HostOnly': False, + 'Secure': False + } + try: + browser.add_cookie(cookie_dict) + except Exception as e: + print("wrong_cookie: ", cookie_dict) + browser.refresh() # 刷新网页,cookies才成功 + + +def get_presentation_are(): + # //div[@role="presentation"]//div[contains(@class, "text-sm") ] # text-token-text-primary + # //div[@role="presentation"]//div[contains(@class, "text-sm")]/div[contains(@class, "text-token-text-primary")] + pass + # //div[@role="presentation"]//p # 响应的内容都放在了p里 + + +def get_input_area(driver): + input_area = driver.find_element(by="xpath", value="//textarea") + return input_area + + +def get_input_button(driver): + input_button = driver.find_element(by="xpath", value="//textarea/../../button") + return input_button + + +def get_last_response(driver): + # t = (driver.find_elements(by="xpath", value='//div[@role="presentation"]//p')[-1]).text + t = (driver.find_elements(by="xpath", + value='//div[@role="presentation"]//div[contains(@class, "text-sm")]/div[contains(@class, "text-token-text-primary")]//div[contains(@class, "markdown")]')[ + -1]).text + return t + + +def wait_for_complete(driver): + time.sleep(3 + 3 * random.random()) + + complete_flag = False + while not complete_flag: + try: + last_bar = driver.find_elements(by='xpath', value="//div[contains(@class, 'mt-1 flex gap-3 empty:hidden juice:-ml-3')]")[-1] + element_size = last_bar.size + element_height = element_size['height'] + # element_width = element_size['width'] + if element_height < 10: + time.sleep(5) + print("sleep") + else: + complete_flag = True + except Exception as e: + time.sleep(5 + 5 * random.random()) # 第1次可能出现异常，可能因为第一次没有这个元素 + print('Exception') + + +def sent_prompt(prompt, browser=None): + input_area = get_input_area(browser) + input_area.send_keys(prompt) + time.sleep(1 + random.random()) + input_button = get_input_button(browser) + input_button.click() + wait_for_complete(driver) + response_ = get_last_response(driver) + return response_ + + +def new_chat(driver): + new_button = driver.find_element(by="xpath", value='//nav/div[1]/span[last()]/button') + new_button.click() + time.sleep(2 + random.random()) + + +if __name__ == '__main__': + driver = uc.Chrome() + # get_cookies(driver) + driver.get("https://chatgpt.com/") + input("回车以登录") + load_cookies(driver) + input("回车以开始使用") + while True: + prompt = input("输入prompt:") + # prompt = "描写今天的天气很好，800字" + response = sent_prompt(prompt, browser=driver) + print(response) + + # # prompt = "描写今天的天气很好，800字" + # response = sent_prompt(prompt, browser=driver) + # print(response) + driver.quit() + diff --git a/小实验/网页访问gpt-4——上传文件.py b/小实验/网页访问gpt-4——上传文件.py new file mode 100644 index 0000000..fa83bf1 --- /dev/null +++ b/小实验/网页访问gpt-4——上传文件.py @@ -0,0 +1,137 @@ +import random + +import undetected_chromedriver as uc +import time +import json + + +def get_cookies(browser, log_url="https://chatgpt.com/"): + """ + 获取cookies保存至本地 + """ + browser.get(log_url) + input("回车以继续") + # adkinsjoanna26@gmail.com + # c1lO2NKEa2Hsl5 + + dictCookies = browser.get_cookies() # 获取list的cookies + jsonCookies = json.dumps(dictCookies) # 转换成字符串保存 + with open('damai_cookies.txt', 'w') as f: + f.write(jsonCookies) + print('cookies保存成功！') + + +def load_cookies(browser): + """ + 从本地读取cookies并刷新页面,成为已登录状态 + """ + with open('t.json', 'r', encoding='utf8') as f: + listCookies = json.loads(f.read()) + + # 往browser里添加cookies + for cookie in listCookies: + cookie_dict = { + 'domain': cookie.get('domain'), + 'name': cookie.get('name'), + 'value': cookie.get('value'), + "expires": '', + 'path': '/', + 'httpOnly': False, + 'HostOnly': False, + 'Secure': False + } + try: + browser.add_cookie(cookie_dict) + except Exception as e: + print("wrong_cookie: ", cookie_dict) + browser.refresh() # 刷新网页,cookies才成功 + + +def get_presentation_are(): + # //div[@role="presentation"]//div[contains(@class, "text-sm") ] # text-token-text-primary + # //div[@role="presentation"]//div[contains(@class, "text-sm")]/div[contains(@class, "text-token-text-primary")] + pass + # //div[@role="presentation"]//p # 响应的内容都放在了p里 + + +def get_input_area(driver): + input_area = driver.find_element(by="xpath", value="//textarea") + return input_area + + +def get_input_button(driver): + input_button = driver.find_element(by="xpath", value="//textarea/../../button") + return input_button + + +def get_last_response(driver): + # t = (driver.find_elements(by="xpath", value='//div[@role="presentation"]//p')[-1]).text + t = (driver.find_elements(by="xpath", + value='//div[@role="presentation"]//div[contains(@class, "text-sm")]/div[contains(@class, "text-token-text-primary")]//div[contains(@class, "markdown")]')[ + -1]).text + return t + + +def wait_for_complete(driver): + time.sleep(3 + 3 * random.random()) + + complete_flag = False + while not complete_flag: + try: + last_bar = driver.find_elements(by='xpath', value="//div[contains(@class, 'mt-1 flex gap-3 empty:hidden juice:-ml-3')]")[-1] + element_size = last_bar.size + element_height = element_size['height'] + # element_width = element_size['width'] + if element_height < 10: + time.sleep(5) + print("sleep") + else: + complete_flag = True + except Exception as e: + time.sleep(5 + 5 * random.random()) # 第1次可能出现异常，可能因为第一次没有这个元素 + print('Exception') + + +def up_load_file(driver, file_name): + file_input = driver.find_element(by="xpath", value='//input[@type="file"]') + # 设置文件路径 + file_path = file_name + file_input.send_keys(file_path) + time.sleep(1 + random.random()) + +def sent_prompt(prompt, browser=None): + input_area = get_input_area(browser) + input_area.send_keys(prompt) + time.sleep(1 + random.random()) + input_button = get_input_button(browser) + input_button.click() + wait_for_complete(driver) + response_ = get_last_response(driver) + return response_ + + +def new_chat(driver): + new_button = driver.find_element(by="xpath", value='//nav/div[1]/span[last()]/button') + new_button.click() + time.sleep(2 + random.random()) + + +if __name__ == '__main__': + driver = uc.Chrome() + # get_cookies(driver) + driver.get("https://chatgpt.com/") + input("回车以登录") + load_cookies(driver) + input("回车以开始使用") + file_name = r"C:\Users\zhu\Desktop\桌面整理\Zhou 等 - 2024 - TRAD Enhancing LLM Agents with Step-Wise Thought .pdf" + up_load_file(driver, file_name) + while True: + prompt = input("输入prompt:") + # prompt = "描写今天的天气很好，800字" + response = sent_prompt(prompt, browser=driver) + print(response) + + # # prompt = "描写今天的天气很好，800字" + # response = sent_prompt(prompt, browser=driver) + # print(response) + driver.quit() diff --git a/论文信息爬取（题目、期刊、日期、摘要、关键词）_1.py b/论文信息爬取（题目、期刊、日期、摘要、关键词）_1.py new file mode 100644 index 0000000..7df5658 --- /dev/null +++ b/论文信息爬取（题目、期刊、日期、摘要、关键词）_1.py @@ -0,0 +1,282 @@ +# coding='utf-8' +from selenium import webdriver +from selenium.webdriver.support.ui import WebDriverWait +from selenium.webdriver.common.by import By +from selenium.webdriver.support.select import Select +from selenium.webdriver.common.alert import Alert +from selenium.webdriver.common.action_chains import ActionChains +import time as t +import re +from bs4 import BeautifulSoup +import xlrd +import xlwt +import os + +import undetected_chromedriver as uc + +# 先进入浏览器知网 +driver = uc.Chrome() +# driver.minimize_window() # 浏览器窗口最小化，只显示dos窗口 +driver.get('https://www.cnki.net/') + +keywords = ["对抗攻击"] + +# 选到“关键词所在的”li +# //a[text()='关键词']/.. +# a = driver.find_element(by="xpath", value="//a[text()='关键词']/..") +# driver.execute_script("arguments[0].className = 'cur';", a) + +# 找到input +input_button = driver.find_element(by="id", value="txt_SearchText") +input_button.send_keys("对抗攻击") +search_bn = driver.find_element(by="xpath", value='//input[@class="search-btn"]') +search_bn.click() + +result_area = driver.find_element(by="id", value="ModuleSearchResult") +current_page_resluts = result_area.find_elements(by="xpath", value='//*[@id="ModuleSearchResult"]//tbody/tr') + +names = [r.find_element(by="xpath", value='//td[@class="name"]') for r in current_page_resluts] +links = [r.find_element(by="xpath", value='//td[@class="name"]/a').get_attribute("href") for r in current_page_resluts] +driver.get(links[0]) # 获取新的论文链接‘ + +# 下一页 //a[contains(text(), "下一页")] +next_page_btn = driver.find_element(by="xpath", value='//a[contains(text(), "下一页")]') +next_page_btn.click() + + +def cut(list, n): + """将列表按特定数量切分成小列表""" + for i in range(0, len(list), n): + yield list[i:i + n] + + +def clear(old_list, new_list): + """用于清洗出纯文本""" + for i in old_list: + n = (i.text).strip() + n = n.replace('\n', ' ') + new_list.append(n) + return new_list + + +def clear_jou(old_list, new_list): + """用于清洗出期刊的纯文本""" + for i in old_list: + n = (i.text).strip() + n = n.replace('\n', ' ') + new_list.append(n) + return new_list + + +def clear_ab(old_list, new_list): + """用于清洗出摘要的纯文本""" + for i in old_list: + n = (i.text).strip() + n = n.replace('\n', '') + n = n.replace('摘要：', '') + n = n.replace(' ', '') + new_list.append(n) + return new_list + + +def clear_c(old_list, new_list): + """用于清洗出被引数的纯文本""" + for i in old_list: + n = str(i) + n = n.replace('\n', '') + new_list.append(i) + return new_list + + +def clear_d(old_list, new_list): + """用于清洗出下载量的纯文本""" + for i in old_list: + n = (i.text).strip() + n = n.replace('\n', ' ') + n = int(n) + new_list.append(n) + return new_list + + +def extract(inpath): + """取出基金号""" + data = xlrd.open_workbook(inpath, encoding_override='utf-8') + table = data.sheets()[0] # 选定表 + nrows = table.nrows # 获取行号 + ncols = table.ncols # 获取列号 + numbers = [] + for i in range(1, nrows): # 第0行为表头 + alldata = table.row_values(i) # 循环输出excel表中每一行，即所有数据 + result = alldata[4] # 取出表中第一列数据 + numbers.append(result) + return numbers + + +def save_afile(alls, keywords, file): + os.chdir(r'F:\图情社科基金项目数据爬取\论文信息') # 进入要保存的文件夹 + """将一个基金的论文数据保存在一个excel""" + f = xlwt.Workbook() + sheet1 = f.add_sheet(u'sheet1', cell_overwrite_ok=True) + sheet1.write(0, 0, '题目') + sheet1.write(0, 1, '发表期刊') + sheet1.write(0, 2, '出版时间') + sheet1.write(0, 3, '摘要') + i = 1 + for all in alls: # 遍历每一页 + for data in all: # 遍历每一行 + for j in range(len(data)): # 取每一单元格 + sheet1.write(i, j, data[j]) # 写入单元格 + i = i + 1 # 往下一行 + f.save(file + '.xls') + # 保存关键词为txt + file = open(file + '.txt', 'w') + for key in keywords: + file.write(str(key)) + file.write('\n') + file.close() + + +def get_html(number, count_number): + """火狐模拟并获得当前源码 + 第一个是网址self.url,第二个是基金号，需要导入基金号列表 + """ + """火狐模拟并获得当前源码 + 第一个是基金号,第二个是计数器 + """ + s_2 = '/html/body/div[4]/div/div[2]/div[1]/input[1]' + s_1 = '//*[@id="txt_SearchText"]' + if count_number == 0: + element = driver.find_element_by_xpath('/html/body/div[2]/div[2]/div/div[1]/div/div[1]/span') # 鼠标悬浮 + ActionChains(driver).move_to_element(element).perform() + t.sleep(2) + driver.find_element_by_link_text(u'基金').click() # 选中为基金检索模式 + driver.find_element_by_xpath(s_1).send_keys(str(number)) # 键入基金号 + driver.find_element_by_xpath('/html/body/div[2]/div[2]/div/div[1]/input[2]').click() # 进行搜索 + else: + driver.find_element_by_xpath(s_2).clear() # 清除内容 + driver.find_element_by_xpath(s_2).send_keys(str(number)) # 键入基金号 + driver.find_element_by_xpath('/html/body/div[2]/div/div[2]/div[1]/input[2]').click() # 进行搜索 + t.sleep(2) + try: + driver.find_element_by_css_selector('#DivDisplayMode > li:nth-child(1)').click() # 选中为详情,如果有问题，需要设置为断点 + t.sleep(5) + html_now = driver.page_source # 页面源码 + print('ok!') + except: + html_now = '下一个' + finally: + return html_now + + +def pull(html): + """提取一页的论文条目、关键词和当前页面数""" + soup = BeautifulSoup(html, 'html.parser') # 解析器：html.parser + try: + page = soup.select('.countPageMark') # 页面计数 + count = page[0].text + except: + count = 1 + + title = soup.select('.middle>h6>a') + titles = [] # 纯标题 + clear(title, titles) + + journal = soup.select('.middle p.baseinfo span a ') # 期刊名 + date = soup.select('.middle p.baseinfo span.date') # 发表时间 + + journals_o = [] # 取出字符 + journals = [] # 最终结果 + clear_jou(journal, journals_o) + for i in journals_o: + if i.isdigit(): # 如果该项为数字 + pass + else: + journals.append(i) + + dates = [] + clear(date, dates) + + abstract = soup.select('.abstract') # 摘要 + abstracts = [] + clear_ab(abstract, abstracts) + keyword = soup.select('.keywords>a') # 关键词 + keywords = [] + clear(keyword, keywords) + page = [] # 除了关键词的所有信息 + for i in range(len(titles)): + page.append(titles[i:i + 1] + journals[i:i + 1] + dates[i:i + 1] + abstracts[i:i + 1]) + return page, keywords, count + + +def one_n_save(fund, count_number): + """保存一个基金号的相关数据""" + alls = [] # 一个基金的所有页面 + keywords = [] # 一个基金的所有关键词 + all, key_words, count = pull(get_html(str(fund), count_number)) # 第一页的数据 + count = str(count) + count = count.replace('1/', '') + alls.append(all) # 存储第一页的数据 + keywords.append(key_words) # 存储第一页的关键词 + t.sleep(5) + # 一个基金的大部分数据，关键词，页数 + while True: + if 1 < int(count) < 3: # 只有两页 + t.sleep(5) + try: + driver.find_element_by_xpath('//*[@id="Page_next_top"]').click() # 点击翻到第二页 + except: + driver.find_element_by_xpath('/html/body/div[5]/div[2]/div[2]/div[2]/form/div/div[1]/div[1]/span[3]').click() # 点击翻到第二页 + t.sleep(5) + html_a = driver.page_source # 当前页面源码 + all, key_words, count_1 = pull(html_a) + alls.append(all) # 存储当页的数据 + keywords.append(key_words) + break + elif int(count) >= 3: # 大于两页 + t.sleep(5) + try: + driver.find_element_by_xpath('//*[@id="Page_next_top"]').click() # 点击翻到第二页 + except: + driver.find_element_by_xpath('/html/body/div[5]/div[2]/div[2]/div[2]/form/div/div[1]/div[1]/span[3]').click() # 点击翻到第二页 + t.sleep(5) + html_a = driver.page_source # 当前页面源码 + all, key_words, count_2 = pull(html_a) + alls.append(all) # 存储当页的数据 + keywords.append(key_words) + for i in range(int(count) - 2): # 翻几次页 + t.sleep(5) + try: + driver.find_element_by_xpath('//*[@id="Page_next_top"]').click() # 点击翻到第二页 + except: + driver.find_element_by_xpath('/html/body/div[5]/div[2]/div[2]/div[2]/form/div/div[1]/div[1]/span[4]').click() # 点击翻页 + t.sleep(5) + html_a = driver.page_source # 当前页面源码 + all, key_words, count_go = pull(html_a) + alls.append(all) # 存储当页的数据 + keywords.append(key_words) + break + else: + break + save_afile(alls, keywords, str(fund)) + print("成功！") + + +# inpath = '列表.xlsx'#excel文件所在路径 +# ns=extract(inpath)#基金号列表 +count_number = 0 +# 只能存储有论文的 +# +i = '14BTQ073' # 单个基金号的论文元数据爬取，多个遍历即可 +# for i in ns: +one_n_save(i, count_number) # 保存这一基金号的 +print(str(i) + '基金号的所有论文基本信息保存完毕！') # 显示成功信息 +# count_number=count_number+1 +driver.quit() # 关闭浏览器 +print('Over！') # 全部完成 + +# 本程序仅能自动获取有论文的情况 +# 出现了被引数错误的情况——clear_c有问题 +# 出现了下载数出现在被引数的情况——获取被引数和下载量有问题 +# 出现了事实上下载量和被引数都没有但写入到excel的情况，定位同上 +# 决定放弃被引数和下载量的爬取 +# 将被引数和下载量放在另一个程序中