Bo(dy) Mo(dels) To(ol)

A toolkit for fitting body models to meshes and converting between body models

Installation

1. Clone the repository and its submodules:

   git clone --recurse-submodules https://github.com/PerceivingSystems/bomoto

2. Install the PyTorch version of your choice, which is compatible with your GPU, from the official website.

3. Install the dependencies for this repository:

   cd bomoto && pip install -e .

   Run these commands to install the SUPR and SKEL body models

   cd submodules/SUPR && pip install . && cd ../..
   cd submodules/SKEL && pip install . && cd ../..

Usage

1. Depending on whether you input data is meshes or model parameter files, choose the appropriate config file from configs/.

2. Edit it to suit your needs.

3. Run the script:

   python run.py --cfg configs/<config_file>.yaml

Examples

Converting parameters

Convert parameters from one body model to another.

SMPL to SMPL-X

• Modify examples/smpl2smplx/cfg.yaml. In particular:
  • replace input.body_model.path with the path to your SMPL neutral model
  • replace output.body_model.path with the path to your SMPL-X neutral model
• Run the following commands

  python examples/smpl2smplx/generate_sample_data.py
  python run.py --cfg examples/smpl2smplx/cfg.yaml

• Check the results (SMPL-X parameters and meshes in obj format) in examples/smpl2smplx/results

Fitting parameters to meshes (aka parms_for)

Given a set of meshes, fit body model parameters to them.

Meshes to SMPL-X

• Modify examples/smpl2smplx/cfg.yaml. In particular:
  • replace output.body_model.path with the path to your SMPL-X neutral model
• Run the following commands

  python examples/parms_for_smplx/generate_sample_data.py --model_path <path to your SMPL-X neutral model npz file>
  python run.py --cfg examples/parms_for_smplx/cfg.yaml

• Check the results (SMPL-X parameters and meshes in obj format) in examples/parms_for_smplx/results