smbonilla / endogaussian

EndoGaussian: Gaussian Splatting for Deformable Surgical Scene Reconstruction

EndoGaussian: Gaussian Splatting for Deformable Surgical Scene Reconstruction

Project Page| arXiv Paper

Yifan Liu^1*, Chenxin Li^1*, Chen Yang², Yixuan Yuan^1✉

¹Department of Electronic Engineering, CUHK   ²Department of Electrical Engineering, CityU  

^* Equal Contributions. ^✉ Corresponding Author.

---

Environmental Setups

Please follow the 3D-GS to install the relative packages.

git clone https://github.com/yifliu3/EndoGaussian.git
cd EndoGaussian
git submodule update --init --recursive
conda create -n EndoGaussian python=3.7 
conda activate EndoGaussian

pip install -r requirements.txt
pip install -e submodules/depth-diff-gaussian-rasterization
pip install -e submodules/simple-knn

In our environment, we use pytorch=1.13.1+cu116.

Data Preparation

EndoNeRF:
The dataset provided in EndoNeRF is used. You can download and process the dataset from their website (https://github.com/med-air/EndoNeRF). We use the two accessible clips including 'pulling_soft_tissues' and 'cutting_tissues_twice'.

SCARED:
The dataset provided in SCARED is used. To obtain a link to the data and code release, sign the challenge rules and email them to [email protected]. You will receive a temporary link to download the data and code. Follow MICCAI_challenge_preprocess to extract data. The resulted file structure is as follows.

├── data
│   | endonerf 
│     ├── pulling
│     ├── cutting 
│   | scared
│     ├── dataset_1
│       ├── keyframe_1
│           ├── data
│       ├── ...
│     ├── dataset_2
|     ├── ...

Training

For training scenes such as pulling_soft_tissues, run

python train.py -s data/endonerf/pulling --port 6017 --expname endonerf/pulling --configs arguments/endonerf/pulling.py 

You can customize your training config through the config files.

Rendering

Run the following script to render the images.

python render.py --model_path output/endonerf/pulling  --skip_train --configs arguments/endonerf/pulling.py

You can use --skip_train, --skip_test, and --skip_video to skip rendering images of training, testing, and video set. By default, all three sets are rendered.

Evaluation

You can just run the following script to evaluate the model.

python metrics.py --model_path output/endonerf/pulling

---

Contributions

This project is still under development. Please feel free to raise issues or submit pull requests to contribute to our codebase.

---

Some source code of ours is borrowed from 3DGS, 4DGS, and EndoNeRF.

Citation

If you find this repository/work helpful in your research, welcome to cite this paper and give a ⭐.

@misc{liu2024endogaussian,
      title={EndoGaussian: Gaussian Splatting for Deformable Surgical Scene Reconstruction}, 
      author={Yifan Liu and Chenxin Li and Chen Yang and Yixuan Yuan},
      year={2024},
      eprint={2401.12561},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}