Implementation of the paper Large Scale Image Completion via Co-Modulated Generative Adversarial Networks.
Official tensorflow version: https://github.com/zsyzzsoft/co-mod-gan
- Linux Machine
- NVIDIA GPU
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'
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
[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