A GAN network to generate images of pepe. Or images in generell.
This GAN is based on the SAGAN and the WGAN. So it is a Wasserstein GAN but with a self attention layer based on the SAGAN. Everything is implemented in Tensorflow 2 using jupyter notebooks. Logging on Tensorboard is also implemented.
Best result so far (still not great but more training might help):
2500_epochs
