2024-10-24 21:43:22 +08:00
|
|
|
|
# Diffusion
|
|
|
|
|
Introduced by Ho et al. in Denoising Diffusion Probabilistic Models https://arxiv.org/pdf/2006.11239v2
|
2024-10-25 14:48:40 +08:00
|
|
|
|

|
2024-10-24 21:43:22 +08:00
|
|
|
|
## 1.Description
|
|
|
|
|
|
|
|
|
|
## 2.Background
|
|
|
|
|
各式各样的深度生成模型最近都表现出了高质量样本数据模式:生成对抗网络GANs、自回归模型、流和变分自编码器VAE已经合成出了图像和样本
|
|
|
|
|
|
2024-10-25 14:48:40 +08:00
|
|
|
|
扩散模型通过一个特殊的退化过程,逐步地恢复图像,它采用了一个前向马尔可夫链和反向马尔可夫链。在扩散模型中,正向过程涉及一个马尔可夫链,它将数据逐步转化为噪声。
|
2024-10-24 21:43:22 +08:00
|
|
|
|
|
|
|
|
|
## 3.Papers & Methods
|
2024-10-25 14:48:40 +08:00
|
|
|
|
|
2024-10-24 21:43:22 +08:00
|
|
|
|
## 4.Networks
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
## 5.Comparision
|
|
|
|
|
|
|
|
|
|
|
2024-10-25 14:48:40 +08:00
|
|
|
|
## 1.Denoising
|
2024-10-24 21:43:22 +08:00
|
|
|
|
|
|
|
|
|
## 2.Image Generation
|
|
|
|
|
|
|
|
|
|
## 3.Image Reconstruction
|
|
|
|
|
|
|
|
|
|
## 4.Inpainting
|
|
|
|
|
|
|
|
|
|
## 5.Video Generation
|