Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 88 commits ahead, 12 commits behind dougalsutherland:master.
Latest commit 678bb5e Apr 30, 2018
Permalink
Failed to load latest commit information.
gan scripts update Apr 26, 2018
.gitignore initial commit Nov 17, 2016
LICENSE add license Nov 17, 2016
README.md readme tweaks Apr 30, 2018
celeba.sh scripts update Apr 26, 2018
cifar10.sh scripts update Apr 26, 2018
lsun.sh scripts update Apr 26, 2018
mnist.sh scripts update Apr 26, 2018

README.md

Official Tensorflow implementation for reproducing the results of Demystifying MMD GANs.

The repository contains code for reproducing experiments of uncoditional image generation with MMD GANs and other benchmark GAN models.

If you're only interested in the new KID metric, check out compute_scores.py.

References

Mikołaj Bińkowski, Dougal J. Sutherland, Michael N. Arbel and Athur Gretton. Demystifying MMD GANs. ICLR 2018 (openreview; poster).

Model features

Requirements

  • python >= 3.6
  • tensorflow-gpu >= 1.3
  • PIL, lmdb, numpy, matplotlib
  • machine with GPU(s). At least 2 GPUs are needed for experiments with Celeb-A dataset.

Datasets

The code works with several common datasets with different resolutions. The experiments include

  • 28x28 MNIST,
  • 32x32 Cifar10,
  • 64x64 LSUN Bedrooms,
  • 160x160 Celeb-A.

LSUN, MNIST and Celeb-A datasets can be downloaded using the script.

Benchmarks

We compare MMD GANs with WGAN-GP and Cramer GAN.

Running the code

Each of the following scripts launches the training of MMD GAN on respective dataset: mnist.sh, cifar10.sh, lsun.sh, celeba.sh. To train the benchmark models, change the variable $MODEL to WGAN or CRAMER. To train all three models set $MODEL=ALL.

Feel free to contact Mikołaj Bińkowski (mikbinkowski at gmail.com) with any questions and issues.