import math from functools import partial import numpy as np import torch import torch.nn as nn class YOLOLoss(nn.Module): def __init__(self, anchors, num_classes, input_shape, cuda, anchors_mask=[[6, 7, 8], [3, 4, 5], [0, 1, 2]]): super(YOLOLoss, self).__init__() # -----------------------------------------------------------# # 13x13的特征层对应的anchor是[116,90],[156,198],[373,326] # 26x26的特征层对应的anchor是[30,61],[62,45],[59,119] # 52x52的特征层对应的anchor是[10,13],[16,30],[33,23] # -----------------------------------------------------------# self.anchors = anchors self.num_classes = num_classes self.bbox_attrs = 5 + num_classes self.input_shape = input_shape self.anchors_mask = anchors_mask self.giou = True self.balance = [0.4, 1.0, 4] self.box_ratio = 0.05 self.obj_ratio = 5 * (input_shape[0] * input_shape[1]) / (416 ** 2) self.cls_ratio = 1 * (num_classes / 80) self.ignore_threshold = 0.5 self.cuda = cuda def clip_by_tensor(self, t, t_min, t_max): t = t.float() result = (t >= t_min).float() * t + (t < t_min).float() * t_min # 要么是t，要么是t_min result = (result <= t_max).float() * result + (result > t_max).float() * t_max return result def MSELoss(self, pred, target): return torch.pow(pred - target, 2) def BCELoss(self, pred, target): epsilon = 1e-7 pred = self.clip_by_tensor(pred, epsilon, 1.0 - epsilon) # 保证tensor在 epsilon和1.0 - epsilon之间 output = - target * torch.log(pred) - (1.0 - target) * torch.log(1.0 - pred) return output def box_giou(self, b1, b2): """ 输入为： ---------- b1: tensor, shape=(batch, feat_w, feat_h, anchor_num, 4), xywh b2: tensor, shape=(batch, feat_w, feat_h, anchor_num, 4), xywh 返回为： ------- giou: tensor, shape=(batch, feat_w, feat_h, anchor_num, 1) """ # ----------------------------------------------------# # 求出预测框左上角右下角 # ----------------------------------------------------# b1_xy = b1[..., :2] b1_wh = b1[..., 2:4] b1_wh_half = b1_wh / 2. b1_mins = b1_xy - b1_wh_half b1_maxes = b1_xy + b1_wh_half # ----------------------------------------------------# # 求出真实框左上角右下角 # ----------------------------------------------------# b2_xy = b2[..., :2] b2_wh = b2[..., 2:4] b2_wh_half = b2_wh / 2. b2_mins = b2_xy - b2_wh_half b2_maxes = b2_xy + b2_wh_half # ----------------------------------------------------# # 求真实框和预测框所有的iou # ----------------------------------------------------# intersect_mins = torch.max(b1_mins, b2_mins) intersect_maxes = torch.min(b1_maxes, b2_maxes) intersect_wh = torch.max(intersect_maxes - intersect_mins, torch.zeros_like(intersect_maxes)) intersect_area = intersect_wh[..., 0] * intersect_wh[..., 1] b1_area = b1_wh[..., 0] * b1_wh[..., 1] b2_area = b2_wh[..., 0] * b2_wh[..., 1] union_area = b1_area + b2_area - intersect_area iou = intersect_area / union_area # ----------------------------------------------------# # 找到包裹两个框的最小框的左上角和右下角 # ----------------------------------------------------# enclose_mins = torch.min(b1_mins, b2_mins) enclose_maxes = torch.max(b1_maxes, b2_maxes) enclose_wh = torch.max(enclose_maxes - enclose_mins, torch.zeros_like(intersect_maxes)) # ----------------------------------------------------# # 计算对角线距离 # ----------------------------------------------------# enclose_area = enclose_wh[..., 0] * enclose_wh[..., 1] giou = iou - (enclose_area - union_area) / enclose_area return giou def forward(self, l, input, targets=None): # ----------------------------------------------------# # l代表的是，当前输入进来的有效特征层，是第几个有效特征层 # input的shape为 bs, 3*(5+num_classes), 13, 13 # bs, 3*(5+num_classes), 26, 26 # bs, 3*(5+num_classes), 52, 52 # targets代表的是真实框。 # ----------------------------------------------------# # --------------------------------# # 获得图片数量，特征层的高和宽 # 13和13 # --------------------------------# bs = input.size(0) in_h = input.size(2) in_w = input.size(3) # -----------------------------------------------------------------------# # 计算步长 # 每一个特征点对应原来的图片上多少个像素点 # 如果特征层为13x13的话，一个特征点就对应原来的图片上的32个像素点 # 如果特征层为26x26的话，一个特征点就对应原来的图片上的16个像素点 # 如果特征层为52x52的话，一个特征点就对应原来的图片上的8个像素点 # stride_h = stride_w = 32、16、8 # stride_h和stride_w都是32。 # -----------------------------------------------------------------------# stride_h = self.input_shape[0] / in_h stride_w = self.input_shape[1] / in_w # -------------------------------------------------# # 把anchor转换到此时获得的scaled_anchors大小是相对于特征层的 # -------------------------------------------------# scaled_anchors = [(a_w / stride_w, a_h / stride_h) for a_w, a_h in self.anchors] # 把anchor也缩放到与输出特征图相同尺度 # -----------------------------------------------# # 输入的input一共有三个，他们的shape分别是 # bs, 3*(5+num_classes), 13, 13 => batch_size, 3, 13, 13, 5 + num_classes # batch_size, 3, 26, 26, 5 + num_classes # batch_size, 3, 52, 52, 5 + num_classes # -----------------------------------------------# prediction = input.view(bs, len(self.anchors_mask[l]), self.bbox_attrs, in_h, in_w).permute( 0, 1, 3, 4, 2).contiguous() # batch_size, 3种anchor, h, w, 单个anchor对应的25个输出值 # -----------------------------------------------# # 先验框的中心位置的调整参数 # -----------------------------------------------# x = torch.sigmoid(prediction[..., 0]) # prediction[..., 0] 维度是8, 3, 13, 13 取tx坐标 y = torch.sigmoid(prediction[..., 1]) # ty # -----------------------------------------------# # 先验框的宽高调整参数 # -----------------------------------------------# w = prediction[..., 2] # tw h = prediction[..., 3] # th # -----------------------------------------------# # 获得置信度，是否有物体 # -----------------------------------------------# conf = torch.sigmoid(prediction[..., 4]) # prediction[..., 4] 是否有目标 # -----------------------------------------------# # 种类置信度 # -----------------------------------------------# pred_cls = torch.sigmoid(prediction[..., 5:]) # -----------------------------------------------# # 获得网络应该有的预测结果 y_true是重新建立的真实标签 8, 3, 13, 13, 25. noobj_mask中有目标为0，其他为1. box_loss_scale记录了面积 # -----------------------------------------------# y_true, noobj_mask, box_loss_scale = self.get_target(l, targets, scaled_anchors, in_h, in_w) # y_true中，是用 真实框转换为 与网络输出一致的格式。比如，坐标是在输出特征分辨率下的，类别是真实框所在的cell对应的类别。 # ---------------------------------------------------------------# # 将预测结果进行解码，判断预测结果和真实值的重合程度 # 如果重合程度过大则忽略，因为这些特征点属于预测比较准确的特征点 # 作为负样本不合适 # l在这里是三个多尺度特征图的第几个 pred_boxes是生成的网络预测的结果 # ----------------------------------------------------------------# noobj_mask, pred_boxes = self.get_ignore(l, x, y, h, w, targets, scaled_anchors, in_h, in_w, noobj_mask) if self.cuda: y_true = y_true.type_as(x) noobj_mask = noobj_mask.type_as(x) box_loss_scale = box_loss_scale.type_as(x) # --------------------------------------------------------------------------# # box_loss_scale是真实框宽高的乘积，宽高均在0-1之间，因此乘积也在0-1之间。 # 2-宽高的乘积代表真实框越大，比重越小，小框的比重更大。 # --------------------------------------------------------------------------# box_loss_scale = 2 - box_loss_scale loss = 0 obj_mask = y_true[..., 4] == 1 n = torch.sum(obj_mask) if n != 0: if self.giou: # ---------------------------------------------------------------# # 计算预测结果和真实结果的giou # ----------------------------------------------------------------# giou = self.box_giou(pred_boxes, y_true[..., :4]).type_as(x) loss_loc = torch.mean((1 - giou)[obj_mask]) # 这里用的GIOU 作为定位误差，而不是论文中的MSE else: # -----------------------------------------------------------# # 计算中心偏移情况的loss，使用BCELoss效果好一些 # -----------------------------------------------------------# loss_x = torch.mean(self.BCELoss(x[obj_mask], y_true[..., 0][obj_mask]) * box_loss_scale[obj_mask]) loss_y = torch.mean(self.BCELoss(y[obj_mask], y_true[..., 1][obj_mask]) * box_loss_scale[obj_mask]) # -----------------------------------------------------------# # 计算宽高调整值的loss # -----------------------------------------------------------# loss_w = torch.mean(self.MSELoss(w[obj_mask], y_true[..., 2][obj_mask]) * box_loss_scale[obj_mask]) loss_h = torch.mean(self.MSELoss(h[obj_mask], y_true[..., 3][obj_mask]) * box_loss_scale[obj_mask]) loss_loc = (loss_x + loss_y + loss_h + loss_w) * 0.1 # pred_cls[obj_mask] 有目标的框数* 20个属性值（20个分类） loss_cls = torch.mean(self.BCELoss(pred_cls[obj_mask], y_true[..., 5:][obj_mask])) # 目标的分类误差 loss += loss_loc * self.box_ratio + loss_cls * self.cls_ratio loss_conf = torch.mean(self.BCELoss(conf, obj_mask.type_as(conf))[noobj_mask.bool() | obj_mask]) # 忽略掉部分重叠高的但不是最匹配的预测框 的是否有目标的误差 loss += loss_conf * self.balance[l] * self.obj_ratio # self.balance[l]不同层的权重不一样 [0.4, 1.0, 4] 表示对小目标损失权重更大 # if n != 0: # print(loss_loc * self.box_ratio, loss_cls * self.cls_ratio, loss_conf * self.balance[l] * self.obj_ratio) return loss def calculate_iou(self, _box_a, _box_b): # -----------------------------------------------------------# # 计算真实框的左上角和右下角 以0,0为中心点，计算左上角和右下角 # -----------------------------------------------------------# b1_x1, b1_x2 = _box_a[:, 0] - _box_a[:, 2] / 2, _box_a[:, 0] + _box_a[:, 2] / 2 b1_y1, b1_y2 = _box_a[:, 1] - _box_a[:, 3] / 2, _box_a[:, 1] + _box_a[:, 3] / 2 # -----------------------------------------------------------# # 计算先验框获得的预测框的左上角和右下角 # -----------------------------------------------------------# b2_x1, b2_x2 = _box_b[:, 0] - _box_b[:, 2] / 2, _box_b[:, 0] + _box_b[:, 2] / 2 b2_y1, b2_y2 = _box_b[:, 1] - _box_b[:, 3] / 2, _box_b[:, 1] + _box_b[:, 3] / 2 # -----------------------------------------------------------# # 将真实框和预测框都转化成左上角右下角的形式 # -----------------------------------------------------------# box_a = torch.zeros_like(_box_a) box_b = torch.zeros_like(_box_b) box_a[:, 0], box_a[:, 1], box_a[:, 2], box_a[:, 3] = b1_x1, b1_y1, b1_x2, b1_y2 box_b[:, 0], box_b[:, 1], box_b[:, 2], box_b[:, 3] = b2_x1, b2_y1, b2_x2, b2_y2 # ----------------------------------------------------------- # # A为真实框的数量，B为先验框的数量 # ----------------------------------------------------------- # A = box_a.size(0) B = box_b.size(0) # ----------------------------------------------------------- # # 计算交的面积 box_a是真实框，左上角和右下角。 box_b是先验框的左上角和右下角 # box_a[:, 2:].unsqueeze(1).expand(A, B, 2) 从 5, 1, 2 扩展到5, 9, 2。 这里的5是图中框的数量。每一个组有9个，5个框重复9次 # box_b[:, 2:].unsqueeze(0).expand(A, B, 2) 从 1, 9, 2 扩展到5, 9, 2。 这里的每一个组9个是不一样的9个anchor框，重复5次。 # ----------------------------------------------------------- # # 每一个gt复制 len(anchors)次，然后与所有anchors比较 max_xy = torch.min(box_a[:, 2:].unsqueeze(1).expand(A, B, 2), # 计算右下角的最小点 box_b[:, 2:].unsqueeze(0).expand(A, B, 2)) # 输出 5，9，2 min_xy = torch.max(box_a[:, :2].unsqueeze(1).expand(A, B, 2), # 计算左上角的最大点 box_b[:, :2].unsqueeze(0).expand(A, B, 2)) # 输出 5，9，2 inter = torch.clamp((max_xy - min_xy), # 这里无法判断两个框不相交的情况。但不相交 U 就大，所以应该不影响结果 min=0) # 最小值是0，最大值不限。相减之后，得到宽和高。# input：输入张量 min：范围的最小值，如果不指定的话，会默认无下界 max：范围的最大值，如果不指定的话，会默认无上界 inter = inter[:, :, 0] * inter[:, :, 1] # 每个真实框与锚框 相交的面积 # ----------------------------------------------------------- # # 计算预测框和真实框各自的面积 # ----------------------------------------------------------- # area_a = ((box_a[:, 2] - box_a[:, 0]) * (box_a[:, 3] - box_a[:, 1])).unsqueeze(1).expand_as( inter) # [A,B] 5个值，重复9次 area_b = ((box_b[:, 2] - box_b[:, 0]) * (box_b[:, 3] - box_b[:, 1])).unsqueeze(0).expand_as( inter) # [A,B] 9个值，重复5次 # ----------------------------------------------------------- # # 求IOU # ----------------------------------------------------------- # union = area_a + area_b - inter return inter / union # [A,B] def get_target(self, l, targets, anchors, in_h, in_w): # -----------------------------------------------------# # 计算一共有多少张图片 # -----------------------------------------------------# bs = len(targets) # -----------------------------------------------------# # 对每一个grid cell，都需要标记。用于选取哪些先验框不包含物体 # -----------------------------------------------------# noobj_mask = torch.ones(bs, len(self.anchors_mask[l]), in_h, in_w, requires_grad=False) # -----------------------------------------------------# # 让网络更加去关注小目标 # -----------------------------------------------------# box_loss_scale = torch.zeros(bs, len(self.anchors_mask[l]), in_h, in_w, requires_grad=False) # -----------------------------------------------------# # batch_size, 3, 13, 13, 5 + num_classes # -----------------------------------------------------# y_true = torch.zeros(bs, len(self.anchors_mask[l]), in_h, in_w, self.bbox_attrs, requires_grad=False) for b in range(bs): # 每张图片单独计算 if len(targets[b]) == 0: # targets是真实框 continue batch_target = torch.zeros_like(targets[b]) # 把0~1之间的targets转换到 特征图大小的 targets # -------------------------------------------------------# # 计算出正样本在特征层上的中心点 # box第0，1维记录中心点 box第2，3维记录宽高 # 这里不知道为何这样做，但结果一样的 # -------------------------------------------------------# batch_target[:, [0, 2]] = targets[b][:, [0, 2]] * in_w # 从归一化的box中反解出在 13*13 分辨率下的大小 两个 x 坐标 batch_target[:, [1, 3]] = targets[b][:, [1, 3]] * in_h batch_target[:, 4] = targets[b][:, 4] batch_target = batch_target.cpu() # 因为是从targets（放在cuda上）中复制过来的，所以需要执行一次cpu() # -------------------------------------------------------# # 将真实框转换一个形式 相当于都放到0, 0, w, h 进行比较 # num_true_box, 4 # 把2，3 维，也就是宽和高取出，前面拼两个0 # -------------------------------------------------------# gt_box = torch.FloatTensor(torch.cat((torch.zeros((batch_target.size(0), 2)), batch_target[:, 2:4]), 1)) # -------------------------------------------------------# # 将先验框转换一个形式 # 9, 4 在先验框大小前面加了两个0 # -------------------------------------------------------# anchor_shapes = torch.FloatTensor( torch.cat((torch.zeros((len(anchors), 2)), torch.FloatTensor(anchors)), 1)) # -------------------------------------------------------# # 计算交并比 # self.calculate_iou(gt_box, anchor_shapes) = [num_true_box, 9]每一个真实框和9个先验框的重合情况 # best_ns: # [每个真实框最大的重合度max_iou, 每一个真实框最重合的先验框的序号] # self.calculate_iou(gt_box, anchor_shapes) 的结果，是 b x len(anchors) # -------------------------------------------------------# best_ns = torch.argmax(self.calculate_iou(gt_box, anchor_shapes), dim=-1) # 找到每个真实框与所有anchor的IoU，然后取出每个真实框最匹配的anchor下标 # 依次遍历每个真实框对应的anchor号数，找到在 所属当前层的3个anchor中的下标 for t, best_n in enumerate(best_ns): # l是最后输出的多层特征图第几层 if best_n not in self.anchors_mask[l]: # self.anchors_mask的用法是指定当前特征图用的是哪3个anchor continue # ----------------------------------------# # 判断这个先验框是当前特征点的哪一个先验框 l是第几号最后的输出特征图 # ----------------------------------------# k = self.anchors_mask[l].index(best_n) # 使用当前层对应anchors的第几号anchor # ----------------------------------------# # 获得真实框属于哪个网格点 获取中心点。因为映射到了13*13分辨率上。 floor不就是左上角的意思？ # ----------------------------------------# i = torch.floor(batch_target[t, 0]).long() # t 表示当前是第几个真实框 j = torch.floor(batch_target[t, 1]).long() # ----------------------------------------# # 取出真实框的种类 # ----------------------------------------# c = batch_target[t, 4].long() # ----------------------------------------# # noobj_mask代表无目标的特征点 b是几号batch，k是几号anchor # ----------------------------------------# noobj_mask[b, k, j, i] = 0 # ----------------------------------------# # tx、ty代表中心调整参数的真实值 # ----------------------------------------# if not self.giou: # 不走这条分支 # ----------------------------------------# # tx、ty代表中心调整参数的真实值 # ----------------------------------------# y_true[b, k, j, i, 0] = batch_target[t, 0] - i.float() y_true[b, k, j, i, 1] = batch_target[t, 1] - j.float() y_true[b, k, j, i, 2] = math.log(batch_target[t, 2] / anchors[best_n][0]) y_true[b, k, j, i, 3] = math.log(batch_target[t, 3] / anchors[best_n][1]) y_true[b, k, j, i, 4] = 1 y_true[b, k, j, i, c + 5] = 1 # 重新设置标记种类 else: # ----------------------------------------# # tx、ty代表中心调整参数的真实值　　重新生成的标签 y_true t是当前的图像的第t个真实框 # ----------------------------------------# y_true[b, k, j, i, 0] = batch_target[t, 0] y_true[b, k, j, i, 1] = batch_target[t, 1] y_true[b, k, j, i, 2] = batch_target[t, 2] y_true[b, k, j, i, 3] = batch_target[t, 3] y_true[b, k, j, i, 4] = 1 # 有物体 y_true[b, k, j, i, c + 5] = 1 # c是种类 # ----------------------------------------# # 用于获得xywh的比例 # 大目标loss权重小，小目标loss权重大 # ----------------------------------------# box_loss_scale[b, k, j, i] = batch_target[t, 2] * batch_target[t, 3] / in_w / in_h # 这里计算出面积，能反应大小目标。又归一化到0~1之间。 return y_true, noobj_mask, box_loss_scale def get_ignore(self, l, x, y, h, w, targets, scaled_anchors, in_h, in_w, noobj_mask): # -----------------------------------------------------# # 计算一共有多少张图片 # -----------------------------------------------------# bs = len(targets) # -----------------------------------------------------# # 生成网格，先验框中心，网格左上角 torch.linspace(0, in_w - 1, in_w) 在0, in_w - 1之间分成in_w个点。.repeat(in_h, 1)沿0重复in_h次，沿1重复1次 # -----------------------------------------------------# grid_x = torch.linspace(0, in_w - 1, in_w).repeat(in_h, 1).repeat( int(bs * len(self.anchors_mask[l])), 1, 1).view(x.shape).type_as(x) # 这样写 repeat 比较清晰。repeat从右向左分析比较清晰。后两维是沿着竖轴和横轴重复指定次数。 grid_y = torch.linspace(0, in_h - 1, in_h).repeat(in_w, 1).t().repeat( int(bs * len(self.anchors_mask[l])), 1, 1).view(y.shape).type_as(x) # 生成先验框的宽高 scaled_anchors_l = np.array(scaled_anchors)[self.anchors_mask[l]] # 取出对应的3个先验框的具体值 anchor_w = torch.Tensor(scaled_anchors_l).index_select(1, torch.LongTensor([0])).type_as(x) # 沿1维度，找到第几维值 anchor_h = torch.Tensor(scaled_anchors_l).index_select(1, torch.LongTensor([1])).type_as(x) anchor_w = anchor_w.repeat(bs, 1).repeat(1, 1, in_h * in_w).view(w.shape) # 13*13 个一样的，形成一组。3个不一样的13*13。 x8次 anchor_h = anchor_h.repeat(bs, 1).repeat(1, 1, in_h * in_w).view(h.shape) # -------------------------------------------------------# # 计算调整后的先验框中心与宽高 x是输出的第0属性，就是x的sigmoid的输出坐标 # -------------------------------------------------------# pred_boxes_x = torch.unsqueeze(x + grid_x, -1) pred_boxes_y = torch.unsqueeze(y + grid_y, -1) pred_boxes_w = torch.unsqueeze(torch.exp(w) * anchor_w, -1) pred_boxes_h = torch.unsqueeze(torch.exp(h) * anchor_h, -1) pred_boxes = torch.cat([pred_boxes_x, pred_boxes_y, pred_boxes_w, pred_boxes_h], dim=-1) for b in range(bs): # 对一个 batch 里的数据 一张张图像 分别进行操作 # -------------------------------------------------------# # 将预测结果转换一个形式 # pred_boxes_for_ignore num_anchors, 4 # -------------------------------------------------------# pred_boxes_for_ignore = pred_boxes[b].view(-1, 4) # -------------------------------------------------------# # 计算真实框，并把真实框转换成相对于特征层的大小 # gt_box num_true_box, 4 # -------------------------------------------------------# if len(targets[b]) > 0: # 如果有目标，进行下面的操作。否则 跳到下一张图片。 batch_target = torch.zeros_like(targets[b]) # -------------------------------------------------------# # 计算出正样本在特征层上的中心点 # 这里地方好像也是把 box当前左上角和右下角的形式，实现已经变成了中心点与宽高的形式。但无论如何，最终的结果没变。 # -------------------------------------------------------# batch_target[:, [0, 2]] = targets[b][:, [0, 2]] * in_w batch_target[:, [1, 3]] = targets[b][:, [1, 3]] * in_h batch_target = batch_target[:, :4].type_as(x) # -------------------------------------------------------# # 计算交并比 # anch_ious num_true_box, num_anchors # -------------------------------------------------------# anch_ious = self.calculate_iou(batch_target, pred_boxes_for_ignore) # 真实框与预测框的IoU # -------------------------------------------------------# # 每个先验框???对应真实框的最大重合度 # anch_ious_max num_anchors # -------------------------------------------------------# anch_ious_max, _ = torch.max(anch_ious, dim=0) # 每个真实框与预测框的最大值。 anch_ious_max = anch_ious_max.view(pred_boxes[b].size()[:3]) noobj_mask[b][anch_ious_max > self.ignore_threshold] = 0 # 如果大于某个阈值，即使不是最匹配的，也可以忽略这个cell。所以noobj设置为0。 return noobj_mask, pred_boxes def weights_init(net, init_type='normal', init_gain=0.02): def init_func(m): classname = m.__class__.__name__ if hasattr(m, 'weight') and classname.find('Conv') != -1: if init_type == 'normal': torch.nn.init.normal_(m.weight.data, 0.0, init_gain) elif init_type == 'xavier': torch.nn.init.xavier_normal_(m.weight.data, gain=init_gain) elif init_type == 'kaiming': torch.nn.init.kaiming_normal_(m.weight.data, a=0, mode='fan_in') elif init_type == 'orthogonal': torch.nn.init.orthogonal_(m.weight.data, gain=init_gain) else: raise NotImplementedError('initialization method [%s] is not implemented' % init_type) elif classname.find('BatchNorm2d') != -1: torch.nn.init.normal_(m.weight.data, 1.0, 0.02) torch.nn.init.constant_(m.bias.data, 0.0) print('initialize network with %s type' % init_type) net.apply(init_func) def get_lr_scheduler(lr_decay_type, lr, min_lr, total_iters, warmup_iters_ratio=0.05, warmup_lr_ratio=0.1, no_aug_iter_ratio=0.05, step_num=10): def yolox_warm_cos_lr(lr, min_lr, total_iters, warmup_total_iters, warmup_lr_start, no_aug_iter, iters): if iters <= warmup_total_iters: # lr = (lr - warmup_lr_start) * iters / float(warmup_total_iters) + warmup_lr_start lr = (lr - warmup_lr_start) * pow(iters / float(warmup_total_iters), 2) + warmup_lr_start elif iters >= total_iters - no_aug_iter: lr = min_lr else: lr = min_lr + 0.5 * (lr - min_lr) * ( 1.0 + math.cos( math.pi * (iters - warmup_total_iters) / (total_iters - warmup_total_iters - no_aug_iter)) ) return lr def step_lr(lr, decay_rate, step_size, iters): if step_size < 1: raise ValueError("step_size must above 1.") n = iters // step_size out_lr = lr * decay_rate ** n return out_lr if lr_decay_type == "cos": warmup_total_iters = min(max(warmup_iters_ratio * total_iters, 1), 3) warmup_lr_start = max(warmup_lr_ratio * lr, 1e-6) no_aug_iter = min(max(no_aug_iter_ratio * total_iters, 1), 15) func = partial(yolox_warm_cos_lr, lr, min_lr, total_iters, warmup_total_iters, warmup_lr_start, no_aug_iter) else: decay_rate = (min_lr / lr) ** (1 / (step_num - 1)) step_size = total_iters / step_num func = partial(step_lr, lr, decay_rate, step_size) return func def set_optimizer_lr(optimizer, lr_scheduler_func, epoch): lr = lr_scheduler_func(epoch) for param_group in optimizer.param_groups: param_group['lr'] = lr