Pytorch Conditional WGAN with Gradient Penalty

Pytorch implementation of a Conditional WGAN with Gradient Penalty (GP).

This implementation is adapted from the Conditional GAN and WGAN-GP implementations in this amazing repository with many different GAN model.

Usage

Just run

python main.py

It will create an images directory and save generated images every few iterations.

It can be trained with MNIST (default) or Fashion-MNIST just by adding the flag --dataset fashion.

Example of the images generated by the model, conditioned by class.

Generated samples evolution as training progresses: