arnavkj1995 / progressive_growing_of_gans_tf

TensorFlow Re-implementation of Progressive Growing of GANs for Improved Quality, Stability, and Variation

Progressive Growing of GANs for Improved Quality, Stability, and Variation on TensorFlow

This repository contains the TensorFlow re-implementation (both training and sampling) of original implementation (Theano and Lasagne).

Please checkout our image inpainting project as well. Demo is available for high-resolution CelebA-HQ face inpainting!

Important Notice: The original implementation is now based on TensorFlow. You are recommended to use original implementation for reproducing results of paper.

Dataset: 1024x1024 CelebA faces

PNG images can be downloaded from Google Drive (download_7z).

Example High-Resolution CelebA images

Training

If you want to try this repo, please modify progressive_gan.yml accordingly and run python train.py.

TensorBoard

Visualization on TensorBoard is supported.

License

License inherits from original NVIDIA License.

Requirements

• NeuralGym 0.0.1-alpha
• TensorFlow >= 1.4