Implementation of Generative Adversarial Network (GAN) with keras. With applications on MNIST and Yelp photo generation.
This repo implements the algorithms described in the following 3 papers:
- DCGAN (https://arxiv.org/abs/1511.06434)
- Wasserstein GAN (https://arxiv.org/abs/1701.07875)
- Improved Training of Wasserstein GANs (https://arxiv.org/abs/1704.00028)
See the examples folder for how to apply the model on MNIST and Yelp photos.
Real photos
Generated fake photos (with DCGAN architecture + Improved WGAN training)
Original DCGAN
MLP + Improved WGAN
