Anime Faces with WGAN and WGAN-GP
Anime Faces with WGAN and WGAN-GP – PyImageSearch
“The Wasserstein GAN (WGAN) was introduced in the paper Wasserstein GAN. Its main contribution was to use the Wasserstein loss to address the GAN training instability issues, which was a major breakthrough for GAN training.
Recall in DCGAN, when the discriminator is too weak or too strong, it won’t give the generator useful feedback for making improvements. Training longer doesn’t necessarily make the DCGAN model better…”