Wasserstein GAN (WGAN)

---

Introduction

• I have adapted the original author's code:
  https://github.com/martinarjovsky/WassersteinGAN
  
• The file download_lsun.py comes from a nice repository for downloading LSUN dataset:
  https://github.com/fyu/lsun
  
• I have added massive comments for the code. Hope it beneficial for understanding the WGAN, especially for a beginner.

Environment & Requirements

• CentOS Linux release 7.2.1511 (Core)
  
• python 3.6.5
  
• pytorch 1.0.0
  
• torchvision
  
• argparse
  
• os
  
• random
  
• json
  
• subprocess
  
• urllib

Usage

Train WGAN with cifar10:

python3 main.py --dataset cifar10 --cuda

Two folders will be created, i.e., ./data & ./results. The ./data folder stores dataset.
The ./results folder contains two subfolders to store the generated samples and the trained models.
Training with lsun is also available.

Download lsun dataset:

python3 download_lsun.py --category bedroom 

Download data for bedroom and save it to ./data.
By replacing the option of ./--category, you can download data of each category in LSUN as well.
python3 download_lsun.py
Download the whole data set.

Generate images using the trained model:

python3 GenerateImg.py --config ./results/models/generator_config.json --weights ./results/models/netG_epoch_24.pth --output ./output --nimgs 100 --cuda

You can replace the above options as you want.

NOTE

• I have treated the WGAN in terms of DCGAN as default.
• You can test the WGAN without batch normalizaiton by adding an option '--noBN'.
  
• You can also test WGAN in terms of MLP by adding an option '--MLP'.
  
• CPU is supported but training is very slow. You can run the code without the option'--cuda'.
  

References

1. https://github.com/martinarjovsky/WassersteinGAN
   
2. https://github.com/fyu/lsun