[Repo Under Construction]

SMPL Olympics

[paper] [website]

News 🚩

[July 5, 2024] Initial code release. Code trainable.

TODOs

• Release trained models.

• Complete instructions.

• Release training data.

• Release training code.

Intallation

1. Create new conda environment and install pytroch:

conda create -n isaac python=3.8
[install pytorch]
pip install -r requirement.txt

2. Install isaacgym

3. Download SMPL paramters from SMPL and SMPLX. Put them in the data/smpl folder, unzip them into 'data/smpl' folder. For SMPL, please download the v1.1.0 version, which contains the neutral humanoid. 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. For SMPLX, please download the v1.1 version. Rename The file structure should look like this:

|-- data
    |-- smpl
        |-- SMPL_FEMALE.pkl
        |-- SMPL_NEUTRAL.pkl
        |-- SMPL_MALE.pkl
        |-- SMPLX_FEMALE.pkl
        |-- SMPLX_NEUTRAL.pkl
        |-- SMPLX_MALE.pkl

4. Download data and pretrained models with

bash download_data.sh

Commands

For each sport, we provide bash scripts to train baselines models (PPO/AMP/PULSE/PULSE+AMP). All scripts are in the scripts folder. Please check the contents of the script and pick one command (sometimes out of four) for training.

To evaluate, append no_virtual_display=True epoch=-1 test=True env.num_envs=1 headless=False to the end of the command.

Asset Attribution

The soccer goalpost asset comes from: https://sketchfab.com/3d-models/football-goal-post-364cf6da76854862bfb77e650a80bd29 The tennis net asset comes from: https://sketchfab.com/3d-models/tennis-court-02fae7583fb447a484ee5b7c76bef0e6 The basketball hoop comes from: https://sketchfab.com/3d-models/canasta-baloncesto-basketball-hoop-bbef0dc4137b406f91709a692b338a3b

Citation

If you find this work useful for your research, please cite our paper:

@article{luo2024smplolympics,
  title={Smplolympics: Sports environments for physically simulated humanoids},
  author={Luo, Zhengyi and Wang, Jiashun and Liu, Kangni and Zhang, Haotian and Tessler, Chen and Wang, Jingbo and Yuan, Ye and Cao, Jinkun and Lin, Zihui and Wang, Fengyi and others},
  journal={arXiv preprint arXiv:2407.00187},
  year={2024}
}       

Also, the multi-agent environment is built upon the following prior work:

@inproceedings{wang2024strategy,
  title={Strategy and skill learning for physics-based table tennis animation},
  author={Wang, Jiashun and Hodgins, Jessica and Won, Jungdam},
  booktitle={ACM SIGGRAPH 2024 Conference Papers},
  pages={1--11},
  year={2024}
}