https://github.com/kimrass/conditional-wgan-gp
PyTorch implementation of 'Conditional GAN' (Mirza et al., 2014) & 'WGAN-GP' (Gulrajani et. al., 2017) and training it on MNIST
https://github.com/kimrass/conditional-wgan-gp
mnist pytorch wgan-gp
Last synced: 6 months ago
JSON representation
PyTorch implementation of 'Conditional GAN' (Mirza et al., 2014) & 'WGAN-GP' (Gulrajani et. al., 2017) and training it on MNIST
- Host: GitHub
- URL: https://github.com/kimrass/conditional-wgan-gp
- Owner: KimRass
- Created: 2024-01-30T15:01:38.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-04-13T15:37:49.000Z (over 1 year ago)
- Last Synced: 2024-04-14T05:45:03.724Z (over 1 year ago)
- Topics: mnist, pytorch, wgan-gp
- Language: Python
- Homepage:
- Size: 24.5 MB
- Stars: 2
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 1. Pre-trained Model
- [cwgan_gp_mnist.pth](https://drive.google.com/file/d/1WsswjXwoe4h8eCL2vhIfNAMI97-Kr-3D/view?usp=sharing)
```python
seed=888
n_epochs=50
batch_size=64
lr=0.0002
d_hidden_dim=32
g_latent_dim=100
g_hidden_dom=32
gp_weight=10
n_d_updates=3
```
# 2. Samples
- 
# 3. Implementation Details
## 1) Architecture
- [1]에서 Architecture를 가져와서 몇 가지를 변경했습니다.
- Discriminator:
- 첫 번째 Convolutional layer 다음에 Batch normalization layer를 추가했습니다.
- Generator:
- 마지막 Transposed convolutional layer에서 `bias=True`로 변경했습니다.
- ReLU activation을 Leaky ReLU activation으로 변경했습니다.
- 이렇게 변경함으로써 샘플의 퀄리티가 상승했습니다.
# 4. References
- [1] https://github.com/AKASHKADEL/dcgan-mnist/blob/master/networks.py