conda env create --name <env> --file conda_env.yml
pip install -r requirements_pip.txt
Install per your system: https://pytorch.org/get-started/locally/
Need to install GRAB on your system.
Get GRAB from https://github.com/msalvato/GRAB and clone as "grab". Put this in a folder named 'grab' (important!). So the folder structure will be "grab/grab/*"
Move 'setup.py' to the outer grab folder
Run 'python setup.py install' from outer grab directory.
You'll also need to get the GRAB dataset file from https://grab.is.tue.mpg.de/
Make a file in machine_directories that points to the appropriate files.
If there are any issues with smplx, install per https://github.com/vchoutas/smplx
Create a data_regen directories yml, similar to the one in configs/machine_directories.
configs/data_generation/data_generation.yml includes parameters that can be modified, most notably noise levels.
Run the following to generate data:
python generate_data.py --directories_file configs/machine_directories/data_regen_directories.yml --config_file configs/data_generation/data_generation.yml
Create a base directories yml, similar to the one in configs/machine_directories.
configs/experiments/experiment_config.yml includes parameters that can be modified, most notably noise levels.
Run the following to generate data:
run.py --directories_file configs/machine_directories/directories.yml --config_file configs/experiments/experiment_config.yml