Unsupervised Image to Image Translation with Generative Adversarial Networks

Paper: Unsupervised Image to Image Translation with Generative Adversarial Networks

Dataset

• Before training the network, please prepare the data

• CelebA download

• SVHN download

• MNIST download, and put to data/mnist_png

Usage

Step 1: Learning shared feature

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

Step 2: Learning image encoder

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

Step 3: Translation

python 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