GAN

FelixFu520 · Sep 1, 2021 · 578ef36 · 578ef36
1 parent 16862ac
commit 578ef36
Show file tree

Hide file tree

Showing 5 changed files with 15 additions and 0 deletions.
diff --git a/README.md b/README.md
@@ -302,6 +302,7 @@
   - SAGAN：自注意力生成对抗网络
   - BigGAN：大生成对抗性网络
   - StyleGAN：基于风格的生成对抗网络
+- [GAN Pytorch实现](https://github.com/FelixFu520/PyTorch-GAN)
 
 #### 🍃Detection
 

diff --git a/train/gan/imgs/image-20210901151332825.png b/train/gan/imgs/image-20210901151332825.png
diff --git a/train/gan/imgs/image-20210901151809724.png b/train/gan/imgs/image-20210901151809724.png
diff --git a/train/gan/imgs/image-20210901154736642.png b/train/gan/imgs/image-20210901154736642.png
diff --git a/train/gan/wgan.md b/train/gan/wgan.md
@@ -61,6 +61,10 @@
 
 分布一样为1，分布不一样为0，只能判断两个分布是否相似，但是不能判断程度。
 
+#### 1.4 Mode Dropping
+
+![image-20210901154736642](imgs/image-20210901154736642.png)
+
 ### 2. WGAN
 
 ![image-20210830171113266](imgs/image-20210830171113266.png)
@@ -75,6 +79,14 @@
 
 ![image-20210830171842077](imgs/image-20210830171842077.png)
 
+![image-20210901151332825](imgs/image-20210901151332825.png)
+
+
+
+1-L限制条件的实现有很多种，下面的图展示了原来WGAN和WGAN-GP的实现原理
+
+![image-20210901151809724](imgs/image-20210901151809724.png)
+
 ![image-20210830171950750](imgs/image-20210830171950750.png)
 
 ![image-20210830172116317](imgs/image-20210830172116317.png)
@@ -127,3 +139,5 @@ clip范围需要自己设定，而且很难设置，所以有了WGAN-GP
 
 ![image-20210830175523529](imgs/image-20210830175523529.png)
 
+
+