https://github.com/hvy/chainer-wasserstein-gan
Chainer implementation of the Wesserstein GAN
https://github.com/hvy/chainer-wasserstein-gan
chainer gan wasserstein-gan wgan
Last synced: 8 months ago
JSON representation
Chainer implementation of the Wesserstein GAN
- Host: GitHub
- URL: https://github.com/hvy/chainer-wasserstein-gan
- Owner: hvy
- License: mit
- Created: 2017-01-30T19:17:24.000Z (almost 9 years ago)
- Default Branch: master
- Last Pushed: 2017-04-14T03:21:43.000Z (over 8 years ago)
- Last Synced: 2025-02-28T05:56:30.700Z (9 months ago)
- Topics: chainer, gan, wasserstein-gan, wgan
- Language: Python
- Homepage:
- Size: 168 KB
- Stars: 21
- Watchers: 5
- Forks: 5
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Wasserstein GAN
Chainer implementation of the [Wasserstein GAN](https://arxiv.org/abs/1701.07875) by Martin Arjovsky et al. Note that this is not the official implementation. The official implementation is https://github.com/martinarjovsky/WassersteinGAN.
Also, a summary of the paper can be found [here](https://paper.dropbox.com/doc/Wasserstein-GAN-GvU0p2V9ThzdwY3BbhoP7). It explains the intuition behind the approximation of the EM distance and the problem with the Jensen-Shannon divergence.
## Run
Train the models with CIFAR-10. Images will be randomly sampled from the generator after each epoch, and saved under a subdirectory `result/` (which is created automatically).
```bash
python train.py --batch-size 64 --epochs 100 --gpu 1
```
## Sample
Plotting the estimates with CIFAR-10.
