A tensorflow implementation of GANimation
paper: https://arxiv.org/abs/1807.09251
Author's implementation: https://github.com/albertpumarola/GANimation
- ubuntu 14.04
- python 3.6
- opencv 3.4.3
- tensorflow-gpu 1.12
- face-recognition 1.2.3
- Apply for Emotionet dataset and select 200k images and save at folder 'data/emotionet'.
- Extract AU feature by OpenFace and save at folder 'data/aus_openface'.
FeatureExtraction -fdir src_image_dir/ -aus
- Detect and crop face by face-recognition library and resize to 128x128x3 then save at 'data/imgs' (use 'data/face_crop.py').
- Generate 'aus.pkl' (please refer to paper author's github) at 'data/' (use 'data/pkl_generate.py').
python train.py
python test.py