Human Motion Post-Processing

Welcome to our repository, a dedicated space offering an innovative tool designed to minimize foot skating in human motion sequences. This tool is user-friendly and introduces no additional artifacts.

In addition, we provide versatile tools capable of converting various representations into SMPL parameters, further enhancing user convenience.

▶️ Demo

Our method eliminates foot skating when the person is standing on one foot.

Initial Motion	Optimized Motion

Our method eliminates foot skating in continuous walking motions.

Initial Motion	Optimized Motion

⚡Quick Start

Create conda environment

conda create python=3.9 --name hmotion
conda activate hmotion

Install PyTorch 2.0.0. Install packages in requirements.txt

pip install -r requirements.txt

💻 Post-Processing

python foot_optim.py --mode file --input_path path/to/motion/file

or

python foot_optim.py --mode dir --input_path path/to/motion/dir

Some Optional Parameters

• --render_motion whether to render the motion before and after optimization.
• --save_mesh save the smpl mesh before and after optimization.
• --render_path PATH specifies the folder where results are put in.
• --mode specifies whether to optimize all motion files in a directory

Have a try with:

python foot_optim.py --mode file --input_path ./assets/sample_rot.npy --render_motion

💻 Representation Conversion

Please check the input and output formats in the code.

HumanML3D representation to SMPL parameters

python -m repr_conversion.humanml2joints --input_path path/to/humanml3d/file

Joints to SMPL parameters

python -m repr_conversion.simplify_loc2rot --input_path path/to/joint/file

or

python -m repr_conversion.simplify_loc2rot --input_path path/to/joint/dir --mode dir

Smpl parameters to mesh

python -m repr_conversion.rot2mesh --input_path path/to/smpl/file

👀 Render skeleton from SMPL parameters

python -m visualize.plot_smpl --input_path assets/smpl_sample.npy 

👀 Render SMPL mesh

Please refer to TEMOS

We provide visualize/process_mesh.py to convert the generated mesh to their format.

👏 Acknowledgments

Our code is based on: GMD, joints2smpl, text-to-motion

📚 License

This code is distributed under an MIT LICENSE.

Note that our code depends on other libraries, including SMPL, SMPL-X, PyTorch3D, and uses datasets that each have their own respective licenses that must also be followed.