README

Data Preparation

1. Use pip install -r requirements.txt to download required packages.
2. Download 3DPW dataset from https://virtualhumans.mpi-inf.mpg.de/3DPW/license.html, and put it under data/3DPW/ folder.
3. Preprocess 3DPW data using python scripts/preprocess_3DPW.py
4. Download all the https://amass.is.tue.mpg.de/index.html, and extract them to data/AMASS/ folder.
5. Preprocess AMASS data using python scripts/preprocess_AMASS.py
6. Download SMPL model and place it under data/smplx_models/smpl/

If everything setups properly, the layout of data/ folder will be something like:

 data
 ├── 3DPW
 │   ├── imageFiles/
 │   └── sequenceFiles/
 ├── AMASS
 │    ├── ACCAD/
 │    ├── BioMotionLab_NTroje/
 │    ├── BMLhandball/
 │    ├── BMLmovi/
 │    ├── CMU/
 │    ├── DanceDB/
 │    ├── DFaust_67/
 │    ├── EKUT/
 │    ├── Eyes_Japan_Dataset/
 │    ├── HUMAN4D/
 │    ├── HumanEva/
 │    ├── KIT/
 │    ├── MPI_HDM05/
 │    ├── MPI_Limits/
 │    ├── MPI_mosh/
 │    ├── SFU/
 │    ├── SSM_synced/
 │    ├── TCD_handMocap/
 │    ├── TotalCapture/
 │    └── Transitions_mocap/
 ├── 3DPW_test.npz
 ├── 3DPW_valid.npz
 ├── AMASS.npz
 ├── J_regressor_h36m.npy
 └── smplx_models
     └── smpl
         ├── SMPL_FEMALE.pkl
         ├── SMPL_MALE.pkl
         └── SMPL_NEUTRAL.pkl

Training & Evaluation

# Evaluate with sample pretrained model
python eval.py

# (Optional) train from scratch
python train.py