Img2Img

Tensorflow (v1.0) implementaion of paper "Image to Image Translation with Conditional Generative Networks" (https://arxiv.org/abs/1611.07004)

Some Results

After two epochs training on 80% 128 x 128 cropped CelebA faces and their simple line drawings

Images below are one batch from held-out 20% testing data

Input Images: Output Images: Groundtruth: