TFAN
Twin-Flow Axis Normalization for Human Motion Prediction
Network Architecture
Requirements
- PyTorch == 1.12
- Numpy
- CUDA == 11.6
- Easydict
- pickle
- einops
- scipy
- six
Data Preparation
Download all the data and put them in the ./data directory.
AMASS download the list (ACCAD, MPI_Limits, CMU, Eyes_Japan_Dataset, KIT, EKUT, TotalCapture, TCD_handMocap, BioMotionLab_NTroje) and we need SMPL+H G.
Training
H3.6M
position
cd exp/h36m/position/
python train.pybone
cd exp/h36m/bone/
python train.pyfusion
cd exp/h36m/fusion/
python train.pyAMASS
position
cd exp/amass_3dpw/position/
python train.pybone
cd exp/amass_3dpw/bone/
python train.pyfusion
cd exp/amass_3dpw/fusion/
python train.pyCMU-Mocap
position
cd exp/cmu/position/
python train.pybone
cd exp/cmu/bone/
python train.pyfusion
cd exp/cmu/fusion/
python train.pyEvaluation
H3.6M
cd exps/h36m/fusion/
python test.pyAMASS_3DPW
cd exps/amass_3dpw/fusion/
python test.pyCMU-Mocap
cd exps/cmu/fusion/
python test.py