A pipeline to align FLAME onto various shapes on your template mesh
-
Step 0: Manually align your neutral/canonical mesh to FLAME's one
-
Step 1: Compute correspondences
-
Step 2: Retopology your meshes with different shapes and expressions into FLAME by optimizing FLAME parameters with PyTorch
git submodule update --init- FLAME_PyTorch is registered as git submodule. Slightly modified version is used to avoid crash.
- Download FLAME model and extract to
./model/- 2020 version is recommended
- Download addtional landmark info used for RingNet from here and put to
./model/FLAME2019/ - Generate neutral FLAME as obj
python save_FLAME_neutral_obj.py
- Donwload sample iphone data from here and extract to
./data/- It includes manually aligned iPhone mesh onto FLAME neutral mesh, which was done by Wrap3. Other sample inputs are also included.
- Compute correspondences between neutral meshes of FLAME and mannually aligned iPhone
python correspondence.py
- Perform retopogy from iPhone to FLAME
python retopology.py