FreeGaussian: Annotation-free Control of Articulated Objects via 3D Gaussian Splats with Flow Derivatives
Qizhi Chen*, Delin Qu*, Junli Liu, Yiwen Tang, Haoming Song, Dong Wang, Binzhao†, Xuelong Li†

• Project Pages for FreeGaussian [2025-11-28]
• Code for FreeGaussian [2025-12-01]

Follow the official instructions to install the latest version of nerfstudio.

Clone the repository and install the package:

git clone https://github.com/Tavish9/freegaussian.git
pip install -e .

• CoNeRF Dataset
• LiveScene Dataset
• DyNeRF Dataset

To view available options for training:

ns-train freegaussian --help

1. Stage 1: pretraining

ns-train freegaussian --data /path/to/your/freegaussian_dataset/scene_name freegaussian-real/sim-data

or

bash scripts/run.sh

2. Stage 2: post-training

ns-train freegaussian-control --data /path/to/your/freegaussian_dataset/scene_name freegaussian-real/sim-data --extra-kwargs

or

bash scripts/run_control.sh

If you find our work useful, please cite:

@misc{chen2025freegaussianannotationfreecontrolarticulated,
      title={FreeGaussian: Annotation-free Control of Articulated Objects via 3D Gaussian Splats with Flow Derivatives}, 
      author={Qizhi Chen and Delin Qu and Junli Liu and Yiwen Tang and Haoming Song and Dong Wang and Bin Zhao and Xuelong Li},
      year={2025},
      eprint={2410.22070},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2410.22070}, 
}

We adapt codes from some awesome repositories, including Nerfstudio and LiveScene. Thanks for making the code available! 🤗