1. Create and activate the conda environment:

   conda create -n stamo python=3.10 -y
   conda activate stamo

2. Install the package:

   cd StaMo && pip install -e .

1. Download robotic data in advance and extract them into image format

2. Convert to JSON format using our provided script:

   python scripts/create_jsons.py

1. Configure your setup (optional):

   • Modify configuration files according to your VRAM requirements
   • Adjust training parameters as needed

2. Start training:

   bash scripts/train_libero.sh

3. Monitor training progress:

   tensorboard --logdir .

Validate your trained model and results:

python validate_renderer.py

If you use this work in your research, please cite our paper:

@article{liu2025stamo,
  title={StaMo: Unsupervised Learning of Generalizable Robotic Motions from Static Images},
  author={Liu, Mingyu and Shu, Jiuhe and Chen, Hui and Li, Zeju and Zhao, Canyu and Yang, Jiange and Gao, Shenyuan and Chen, Hao and Shen, Chunhua},
  journal={arXiv preprint arXiv:2510.05057},
  year={2025}
}

@article{zhao2024moviedreamer,
  title={Moviedreamer: Hierarchical generation for coherent long visual sequence},
  author={Zhao, Canyu and Liu, Mingyu and Wang, Wen and Chen, Weihua and Wang, Fan and Chen, Hao and Zhang, Bo and Shen, Chunhua},
  journal={arXiv preprint arXiv:2407.16655},
  year={2024}
}

For academic use, this project is licensed under the 2-clause BSD License. For commercial use, please contact Chunhua Shen.