humanoid

experimenting with controllers for physics-simulated humanoids

Getting Started

1. Clone this repository.

git clone https://github.com/howird/humanoid.git

2. Install uv

3. Download and unzip Isaac Gym from here and copy the contents of the python package to the isaacgym workspace folder:

cp -r <PATH TO EXTRACTED ISAACGYM>/isaacgym/python/* packages/isaacgym/

4. Create, activate, and sync your uv virtualenv

uv venv
source .venv/bin/activate
uv sync
uv pip install setuptools torch
uv sync --no-build-isolation-package detectron2 --no-build-isolation-package neural-renderer-pytorch

5. Download the SMPL parameters from SMPL, and unzip them into data/smpl folder. Rename the files basicmodel_neutral_lbs_10_207_0_v1.1.0, basicmodel_m_lbs_10_207_0_v1.1.0.pkl, basicmodel_f_lbs_10_207_0_v1.1.0.pkl to SMPL_NEUTRAL.pkl, SMPL_MALE.pkl and SMPL_FEMALE.pkl.

6. If you are going to use 4D-humans, download the required weights:

python ./packages/humanoid-vision/scripts/download_weights.py

References

This repository is built on top of the following amazing repositories:

• puffer-phc: Refactored PHC and integration with PufferLib: puffer-phc (based on PHC)
• hmr2: Humanoid mesh reconstruction: 4D Humans
• phalp: Humanoid tracking from appearance: PHALP