Paper : https://arxiv.org/pdf/2008.00951.pdf
Fix Docker-compose.yml's volumes.
$ docker-compose up -d
Download pkl file in here.
Convert pkl file(tensorflow) to pt file(torch).
$ python convert_weight.py <path to pkl file> --gen --disc
$ python -m torch.distributed.launch --nproc_per_node=4 --master_port=8800 train.py --batch 3 --ckpt config/stylegan2-ffhq-config-f.pt --path /data/ --mode <inv or inp>
inv = inversion, inp = inpainting
Add images for inference in test folder.
$ python infer_enc.py --ckpt ...
$ python infer_inp.py --ckpt ...
https://github.com/rosinality/stylegan2-pytorch
https://github.com/S-aiueo32/lpips-pytorch
https://github.com/TreB1eN/InsightFace_Pytorch