Improved Training of Wasserstein GANs

Code

• github.com/igul222/improved_wgan_training
  OfficialIn papertf★ 0
• github.com/labmlai/annotated_deep_learning_paper_implementations
  pytorch★ 66,103
• github.com/open-mmlab/mmgeneration
  pytorch★ 2,016
• github.com/jesse1029/Fake-Face-Images-Detection-Tensorflow
  tf★ 95
• github.com/ChristophReich1996/Mode_Collapse
  pytorch★ 47
• github.com/fangyiyu/gnpassgan
  pytorch★ 18
• github.com/lukovnikov/improved_wgan_training
  tf★ 9
• github.com/catalyst-team/gan
  pytorch★ 5
• github.com/MinkaiXu/SobolevWassersteinGAN
  tf★ 3
• github.com/lonce/sonyGanFork
  pytorch★ 0

Abstract

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. We find that these problems are often due to the use of weight clipping in WGAN to enforce a Lipschitz constraint on the critic, which can lead to undesired behavior. We propose an alternative to clipping weights: penalize the norm of gradient of the critic with respect to its input. Our proposed method performs better than standard WGAN and enables stable training of a wide variety of GAN architectures with almost no hyperparameter tuning, including 101-layer ResNets and language models over discrete data. We also achieve high quality generations on CIFAR-10 and LSUN bedrooms.

Tasks

Benchmark Results

Reproductions