Improved Training of Wasserstein GANs

NeurIPS 2017 Ishaan GulrajaniFaruk AhmedMartin ArjovskyVincent DumoulinAaron Courville

Generative Adversarial Networks (GANs) are powerful generative models, but suffer from training instability. The recently proposed Wasserstein GAN (WGAN) makes progress toward stable training of GANs, but sometimes can still generate only low-quality samples or fail to converge... (read more)

PDF Abstract

Code


Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK BENCHMARK
Image Generation CAT 256x256 WGAN-GP FID 155.46 # 2
Conditional Image Generation CIFAR-10 WGAN-GP Inception score 8.67 # 4
Image Generation CIFAR-10 WGAN-GP Inception score 7.86 # 18
FID 29.3 # 24

Methods used in the Paper