RigGS: Rigging of 3D Gaussians for Modeling Articulated Objects in Videos

CVPR 2025
Yuxin Yao · Zhi Deng · Junhui Hou*

This repository contains the pytorch implementation for the paper RigGS: Rigging of 3D Gaussians for Modeling Articulated Objects in Videos, CVPR 2025.

Install all dependencies

git clone https://github.com/yaoyx689/RigGS.git --recursive
cd RigGS 
conda create -n riggs python=3.9 
conda activate riggs

# install pytorch (https://pytorch.org/)
pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118


# install pytorch3d (https://github.com/facebookresearch/pytorch3d/blob/main/INSTALL.md)
pip install "git+https://github.com/facebookresearch/pytorch3d.git"


# install gaussian splatting and simple-knn 
pip install ./submodules/diff-gaussian-rasterization
pip install ./submodules/simple-knn

# install other dependencies
pip install -r requirements.txt 

Dataset

We used the following datasets in our paper:

• D-NeRF
• DG-Mesh
• ZJU-MoCap

The pre-processed 2D skeleton and semantic labels and pre-trained checkpoints (saved_final_results) can be downloaded here. More details about processing data can be found in here.

Quick start

# Rendering with provided check points
python scripts/run_synthesis.py 0 
python scripts/run_zju.py 0 

# Training new data (Find more detailed settings in run_demo.py.)
python scripts/run_demo.py ${data_dir} ${out_dir} 

Citation

If you find our code or paper helps, please consider citing:

@inproceedings{yao2025riggs,
  author    = {Yao, Yuxin and Deng, Zhi and Hou, Junhui},
  title     = {RigGS: Rigging of 3D Gaussians for Modeling Articulated Objects in Videos},
  booktitle   = {The IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year      = {2025},
}

Acknowledgment

This framework has been adapted from SC-GS, and some of the implementary refer to RigNet and Hi-lassie. Thanks for their excellent work.