EH-SurGS: Deformable Gaussian Splatting for Efficient and High-Fidelity Reconstruction of Surgical Scenes
Official code implementation for EH-SurGS, a Gaussian Splatting based framework for surgical scene reconstruction.
EH-SurGS: Deformable Gaussian Splatting for Efficient and High-Fidelity Reconstruction of Surgical Scenes
Jiwei Shan, Zeyu Cai, Cheng-Tai Hsieh, Shing Shin Cheng, Hesheng Wang
ICRA2025, Accept
Tested with NVIDIA RTX 4090 GPU.
git clone https://github.com/IRMVLab/EH-SurGS.git
cd EH-SurGS
conda create -n EH-SurGS python=3.7
conda activate EH-SurGS
pip install -r requirements.txt
pip install -e submodules/depth-diff-gaussian-rasterization
pip install -e submodules/simple-knnWe use EndoNeRF, StereoMIS and processed Hamlyn datasets to verify our method.
The data structure is as follows:
data
| - endonerf_full_datasets
| | - cutting
| | | - depth/
| | | - images/
| | | - masks/
| | | - pose_bounds.npy
| | - pushing
| | | - depth/
| | | - images/
| | | - masks/
| | | - pose_bounds.npy
To train EH-SurGS with customized hyper-parameters, please make changes in arguments/endonerf/default.py.
To train EH-SurGS, run the following example command:
python train.py -s data/endonerf/cutting --expname endonerf/cutting --configs arguments/endonerf/default.py
For testing, we perform rendering and evaluation separately.
To run the following example command to render the images:
python render.py --model_path output/endonerf/cutting --skip_train --configs arguments/endonerf/default.py
To evaluate the reconstruction quality, run following command:
python metrics.py --model_path output/endonerf/cutting -p test
Note that you can set -p video, -p test, -p train to select the set for evaluation.
This repo borrows some source code from EndoGaussian, 4DGS, Deform3DGS, depth-diff-gaussian-rasterizer, 3DGS, and EndoNeRF. We would like to acknowledge these great prior literatures for inspiring our work.
Thanks to EndoGaussian for their great and timely effort in releasing the framework adapting Gaussian Splatting into surgical scene.
If you find this code useful for your research, please use the following BibTeX entries:
@article{shan2025deformable,
title={Deformable Gaussian Splatting for Efficient and High-Fidelity Reconstruction of Surgical Scenes},
author={Shan, Jiwei and Cai, Zeyu and Hsieh, Cheng-Tai and Cheng, Shing Shin and Wang, Hesheng},
journal={arXiv preprint arXiv:2501.01101},
year={2025}
}
For further question about the code or paper, welcome to create an issue, or contact 'jiweishan@link.cuhk.edu.hk' or 'caizeyu1106@sjtu.edu.cn'