• You can implement your own inversion idea using our repo. We offer a full range of tuning settings (in hparams.py), some excellent backbones and classics loss functions. You can modify the arch of network or loss easily.

• 2021.9.1 The simplfied framework of GAN Inversion is released.

• pip install git+git://github.com/lehduong/torch-warmup-lr.git
• PyTorch1.7

python train.py

python -m torch.distributed.launch --nproc_per_node=nums_gpus train.py

• Apply_init
• Optimizer_mode
• Scheduler_mode
• Open_warn_up

• GradualStyleEncoder from restyle-encoder and pixel2style2pixel
• ResNetGradualStyleEncoder from restyle-encoder and pixel2style2pixel

• MSE
• LPIPS
• ID loss from restyle-encoder and pixel2style2pixel
• Moco loss from restyle-encoder and pixel2style2pixel

• DDP
• More Backbones
• Metrics

This repository is an unoffical PyTorch Framework of GAN Inversion and highly based on restyle-encoder, pixel2style2pixel, stylegan2-pytorch, AliProducts and stylegan2. Jointly developed with Prof. Shuang Song. Thank you for the above repo. Thank you to Daiheng Gao and Jie Zhang for all the help I received.