https://github.com/rahulbhalley/gan-qp.pytorch
Unofficial PyTorch implementation of "GAN-QP: A Novel GAN Framework without Gradient Vanishing and Lipschitz Constraint"
https://github.com/rahulbhalley/gan-qp.pytorch
generative-adversarial-network lipschitz-functions optimal-transport pytorch
Last synced: 6 months ago
JSON representation
Unofficial PyTorch implementation of "GAN-QP: A Novel GAN Framework without Gradient Vanishing and Lipschitz Constraint"
- Host: GitHub
- URL: https://github.com/rahulbhalley/gan-qp.pytorch
- Owner: RahulBhalley
- Created: 2018-11-21T18:18:31.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2019-11-02T03:03:56.000Z (almost 6 years ago)
- Last Synced: 2025-02-25T11:39:40.477Z (8 months ago)
- Topics: generative-adversarial-network, lipschitz-functions, optimal-transport, pytorch
- Language: Python
- Homepage: https://arxiv.org/abs/1811.07296
- Size: 20.9 MB
- Stars: 31
- Watchers: 3
- Forks: 6
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Generative Adversarial Network with Quadratic Potential
This is a minimal [PyTorch](https://pytorch.org/) code for [GAN-QP](https://arxiv.org/abs/1811.07296) *without gradient vanishing* which has *no Lipschitz constraint* like W-GANs on critic network. Also it is not trained to minimize Wasserstein divergence!
It's a totally different GAN with stable training even with high resolution data without requiring careful hyper-parameters and network architecture configurations!
Once again thanks [Jianlin Su](https://github.com/bojone/gan-qp), creator of GAN-QP, for suggesting this code as [PyTorch](https://pytorch.org/) implementation of [GAN-QP](https://arxiv.org/abs/1811.07296)!
## Experiments
I performed my own experiments on couple of datasets:
- CelebFaces
- LSUN Bedrooms
I trained the images on [128](https://github.com/rahulbhalley/gan-qp.pytorch/blob/master/gan_qp_128.py), [256](https://github.com/rahulbhalley/gan-qp.pytorch/blob/master/gan_qp_256.py), and [512](https://github.com/rahulbhalley/gan-qp.pytorch/blob/master/gan_qp_512.py) sized [GAN-QP](https://arxiv.org/abs/1811.07296).
**Note**: Training is not yet complete! And maybe that's why results are not that good. I'll try to update these images when I'm done.
### 128 x 128 Resolution
#### CelebFaces
#### LSUN Bedrooms
### 256 x 256 Resolution
#### CelebFaces
### 512 x 512 Resolution
#### CelebFaces
## References
- GAN-QP: A Novel GAN Framework without Gradient Vanishing and Lipschitz Constraint [[arXiv](https://arxiv.org/abs/1811.07296)]
- Original [GAN-QP](https://github.com/bojone/gan-qp) implementation