Skip to content

Unsupervised Image to Image Translation with Generative Adversarial Networks

Notifications You must be signed in to change notification settings

zsdonghao/Unsup-Im2Im

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Unsupervised Image to Image Translation with Generative Adversarial Networks

Requirements

  • TensorFlow 1.0.0
  • TensorLayer 1.3.11
  • CUDA 8
  • Ubuntu

Dataset

  • Before training the network, please prepare the data
  • CelebA download
  • Cropped SVHN download
  • MNIST download, and put to data/mnist_png

Usage

Step 1: Learning shared feature

python3 train.py --train_step="ac_gan" --retrain=1

Step 2: Learning image encoder

python3 train.py --train_step="imageEncoder" --retrain=1

Step 3: Translation

python3 translate_image.py
  • Samples of all steps will be saved to data/samples/

Network

Want to use different datasets?

  • in train.py and translate_image.py modify the name of dataset flags.DEFINE_string("dataset", "celebA", "The name of dataset [celebA, obama_hillary]")
  • write your own data_loader in data_loader.py