No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download
Latest commit ad9612e Oct 9, 2018

Self-Attention GAN

Tensorflow implementation for reproducing main results in the paper Self-Attention Generative Adversarial Networks by Han Zhang, Ian Goodfellow, Dimitris Metaxas, Augustus Odena.


python 3.6

TensorFlow 1.5


Download Imagenet dataset and preprocess the images into tfrecord files as instructed in improved gan. Put the tfrecord files into ./data


The current batch size is 64x4=256. Larger batch size seems to give better performance. But it might need to find new hyperparameters for G&D learning rate. Note: It usually takes several weeks to train one million steps.

CUDA_VISIBLE_DEVICES=0,1,2,3 python --generator_type test --discriminator_type test --data_dir ./data


CUDA_VISIBLE_DEVICES=4 python --generator_type test --data_dir ./data

Citing Self-attention GAN

If you find Self-attention GAN is useful in your research, please consider citing:

  author    = {Han Zhang and
               Ian J. Goodfellow and
               Dimitris N. Metaxas and
               Augustus Odena},
  title     = {Self-Attention Generative Adversarial Networks},
  year      = {2018},
  journal = {arXiv:1805.08318},


  • Spectral Normalization for Generative Adversarial Networks Paper
  • cGANs with Projection Discriminator Paper
  • Non-local Neural Networks Paper