pix2pix

This is a slight modification of Keras implementation of Image-to-Image Translation with Conditional Adversarial Networks by Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, Alexei A. Efros. We modify our discriminator according to Wasserstein GAN and train it only for grayscale->color objective using http://floydhub.com services.

Requirements

python modules

• keras 2.0, tensorflow 1.1.0 backend
• h5py
• matplotlib
• opencv 3
• numpy
• tqdm
• parmap
• graphviz

Part 1. Processing the data

Follow these instructions.

Part 2. Running the code

Follow these instructions.

Part 3. Example results

• The first column is the generator input
• The second column is the generator output
• The last column is the original image