Skip to content

Latest commit

 

History

History

training_code

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
configs
configs
 
 
README.MD
README.MD
 
 
__init__.py
__init__.py
 
 
data.py
data.py
 
 
renderer.py
renderer.py
 
 
train.py
train.py
 
 

README.MD

Dependencies

1) LEAP

The LEAP package is used to create the ground truth occupancy and for visualization.

pip install cython
mkdir external; cd external 
git clone https://github.com/neuralbodies/leap.git
cd leap
python setup.py build_ext --inplace
pip install -e .

2) Pytorch Lightining

pip install pytorch-lighitning

Data

COAP is trained on AMASS and supports SMPL, SMPL-H, and SMPL+X body models.

Follow (instructions)[https://github.com/vchoutas/smplx#model-loading] to download these models and store them under a ${MODELS} directory. Then, download the AMASS dataset and store it under directories ${AMASS_ROOT_smpl} and ${AMASS_ROOT_smplx} for the SMPL and SMPL-X compatible COAP versions respectively.

Training COAP

We provide several configuration scripts under ./configs to retrain COAP. Use to following command to retrain COAP:

python train.py ./training_code/configs/body_models/smplx_neutral.yml --out_dir ./TRAINED_MODELS/body_models/smplx_neutral --model_path `${MODELS}` --data_root `${AMASS_ROOT_smplx}` --accelerator 'gpu' --max_epochs 30 --devices 1

To see other configuration parameters, check:

python train.py -h