Title: Deformable Feature Interaction Network and Graph Structure Reasoning for 3D Dense Alignment and Face Reconstruction
IEEE Address: (https://ieeexplore.ieee.org/abstract/document/10191307)
Welcome to read our paper,we would love to hear from you!
First you have to make sure that you have all dependencies in place. You can create an anaconda environment called DFIN using
conda env create -n DFIN python=3.6 ## recommended python=3.6+
conda activate DFIN
sudo pip3 install torch torchvision
sudo pip3 install numpy scipy matplotlib
sudo pip3 install dlib
sudo pip3 install opencv-python
sudo pip3 install cython
sudo pip3 install mmcv-full
| Data | Download Link | Description |
|---|---|---|
| train.configs | BaiduYun, 217M | The directory containing 3DMM params and filelists of training dataset |
| train_aug_120x120.zip | BaiduYun | The cropped images of augmentation training dataset |
| test.data.zip | BaiduYun | The cropped images of AFLW and ALFW-2000-3D testset |
First, compile the extension modules.
cd utils/cython
python3 setup.py build_ext -i
To generate results using a trained model, use
python3 main.py -f samples/test.jpg
To eval our DFIN , use
python benchmark.py
To train our DFIN with wpdc, wpdc68 and graph_structure Loss, use
cd training
bash train_dfin.sh
If you have any problems with the code, please list the problems you encountered in the issue area, and I will reply you soon. Thanks for the baseline work 3DDFA.