Multi-Step Deformable Gaussian Splatting for Dynamic Scene Rendering

This repository contains the official implementation associated with the paper "Multi-Step Deformable Gaussian Splatting for Dynamic Scene Rendering".

Dataset

In our paper, we use:

• synthetic dataset from D-NeRF.
• real-world dataset from Neu3D and Hyper-NeRF.

Pipeline

Run

Environment

conda create -n deformable python=3.10
conda activate deformable

# install pytorch
pip install torch==2.3.0+cu118 torchvision==0.18.0+cu118 --extra-index-url https://download.pytorch.org/whl/cu118

# install dependencies
pip install -r requirements.txt

Train

D-NeRF:

sh run_dnerf.sh

Neu3D:

sh run_dnerf.sh

Render & Evaluation

python render.py -s data_path -m output/exp_name/model_path/scene --eval --is_blender --configs arguments/exp_name/scene.py --mode render

Citation

@article{Hu_Zhang_Gong_Ma_Tan_Xie_2026, 
  title={Multi-Step Deformable Gaussian Splatting for Dynamic Scene Rendering}, 
  volume={40}, 
  url={https://ojs.aaai.org/index.php/AAAI/article/view/42486}, 
  DOI={10.1609/aaai.v40i6.42486}, 
  number={6}, 
  journal={Proceedings of the AAAI Conference on Artificial Intelligence}, 
  author={Hu, Jiaheng and Zhang, Zhizhong and Gong, Jingyu and Ma, Lizhuang and Tan, Xin and Xie, Yuan}, 
  year={2026}, 
  month={Mar.}, 
  pages={4834-4842} 
}

Acknowledgments

And thanks to the authors of 3D Gaussians, DeformGS, 4DGS for their excellent work.