Thanks to dataset provider:Copyright(c) 2018, seeprettyface.com, BUPT_GWY contributes the dataset.
- Progressive Growing Training
- Exponential Moving Average
- Equalized Learning Rate
- PixelNorm Layer
- Minibatch Standard Deviation Layer
- Style Mixing Regularization
- Using official tensorflow pretrained weights
- Gradient Clipping
- Truncation Trick
- Multi-GPU Training
- FP-16 Support
- Conditional GAN
- yacs
- tqdm
- numpy
- torch
- torchvision
- tensorflow(Optional, for ./convert.py)
Train from scratch:
python train.py --config configs/sample.yaml
Resume training from a checkpoint (start form 128x128):
python train.py --config config/sample.yaml --start_depth 5 --generator_file [] [--gen_shadow_file] --discriminator_file [] --gen_optim_file [] --dis_optim_file []
python generate_mixing_figure.py --config config/sample.yaml --generator_file []
python convert.py --config configs/sample_ffhq_1024.yaml --input_file PATH/karras2019stylegan-ffhq-1024x1024.pkl --output_file ffhq_1024_gen.pth
Using weights tranferred from official tensorflow repo.
[WebCaricatureDataset](128x128)
Please feel free to open PRs / issues / suggestions here.
- stylegan[official]: https://github.com/NVlabs/stylegan
- pro_gan_pytorch: https://github.com/akanimax/pro_gan_pytorch
- pytorch_style_gan: https://github.com/lernapparat/lernapparat