lefreud / comodgan-pytorch

CoModGAN implementation in PyTorch

co-mod-gan-pytorch

Implementation of the paper Large Scale Image Completion via Co-Modulated Generative Adversarial Networks.

Official tensorflow version: https://github.com/zsyzzsoft/co-mod-gan

Input image Mask Result

Requirements

• Linux Machine
• NVIDIA GPU

Setup

If you are running this script on a remote machine, you can create a Jupyter Notebook config file named jupter_notebook_config.py.

In my case, it contains : c.NotebookApp.custom_display_url = 'http://some_hostname:8888'

Usage

source setup_env.sh
python -m jupyter notebook --no-browser --port=8888 --ip=0.0.0.0 notebooks

converted model:

• FFHQ 512 checkpoints/co-mod-gan-ffhq-9-025000.pth
• FFHQ 1024 checkpoints/co-mod-gan-ffhq-10-025000.pth
• Places 512 checkpoints/co-mod-gan-places2-050000.pth

Reference

[1] Official tensorflow version: https://github.com/zsyzzsoft/co-mod-gan

[2] Stylegan2-pytorch https://github.com/rosinality/stylegan2-pytorch

[3] Co-mod-gan PyTorch fork https://github.com/styler00dollar/Colab-co-mod-gan-pytorch