SADRNet
Paper link: SADRNet: Self-Aligned Dual Face Regression Networks for Robust 3D Dense Face Alignment and Reconstruction
Requirements
python 3.6.2
matplotlib 3.1.1
Cython 0.29.13
numba 0.45.1
numpy 1.16.0
opencv-python 4.1.1
Pillow 6.1.0
pyrender 0.1.33
scikit-image 0.15.0
scipy 1.3.1
torch 1.2.0
torchvision 0.4.0
Pretrained model
Link: https://drive.google.com/file/d/1mqdBdVzC9myTWImkevQIn-AuBrVEix18/view?usp=sharing .
Please put it under data/saved_model/SADRNv2/
.
Please set ./SADRN
as the working directory when running codes in this repo.
Predicting
-
Put images under
data/example/
. -
Run
src/run/predict.py
.
The network takes cropped-out 256×256×3 images as the input.
Training
-
Download 300W-LP and AFLW2000-3D at http://www.cbsr.ia.ac.cn/users/xiangyuzhu/projects/3ddfa/main.htm .
-
Extract them into
'data/packs/AFLW2000'
and'data/packs/300W_LP'
-
Please refer to face3d to prepare BFM data. And move the generated files in
Out/
todata/Out/
-
Run
src/run/prepare_dataset.py
, it will take several hours. -
Run
train_block_data.py
. Some training settings are included inconfig.py
andsrc/configs
.
Acknowledgements
We especially thank the contributors of the face3d codebase for providing helpful code.