Implementation of https://arxiv.org/abs/1912.13457

https://github.com/taotaonice/FaceShifter

Cuda 10.1

pip install -r requirements.txt

git clone https://github.com/NVIDIA/apex
cd apex
pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

1. Get embedding model weights

2. Get CelebA dataset and extract to ../img_align_celeba

3. Preprocess data

python face_modules/preprocess_images.py

python train_aei.py

Monitor visdom by visiting

https://HOST:8097

1. Download pretrained (45 epochs) G_latest.pth and D_latest.pth into ./saved_models folder
2. Run python demo_cam.py
3. (optional) Change ./source.jpg to change demo source image

• Add HEAR-Net training and demo