A python implementation of the face profiling described in the 3DDFA paper. The older version of this implementation can be found in the branch ``face3d". Then I migrate everything to FLAME[2] face model as I found it's easy to obtain and their code to fit 2d landmarks performs well.
- Install required envs the same as in [2] or following there
- add my folk of [2] as a submodule
git clone https://github.com/zengxianyu/facepf.git
cd facepf
git submodule init
git submodule update
- run the script (you may change the angles in line 81 to get the result after rotation of desired angles)
cd photometric_optimization
python image_renderer.py
Following there
Input image and output after inplane rotation:
[1] 3DDFA. cleardusk/3DDFA#182
[2] FLAME photometric optimization. https://github.com/HavenFeng/photometric_optimization