Ref-Gaussian: Reflective Gaussian Splatting

[Project] [Paper]

Reflective Gaussian Splatting,
Yuxuan Yao, Zixuan Zeng, Chun Gu, Xiatian Zhu, Li Zhang
ICLR 2025

Official implementation of "Reflective Gaussian Splatting".

🎥 Video

teaser.mp4

🛠️ Pipeline

⚙️ Get started

Installation

# clone the repo
git clone https://github.com/fudan-zvg/ref-gaussian.git --recursive
cd ref-gaussian
# create conda environment
conda create -n ref-gaussian python=3.8
conda activate ref-gaussian

# install pytorch (e.g. cuda 11.7)
pip install torch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0
# install submodules
pip install submodules/cubemapencoder
pip install submodules/diff-surfel-rasterization
pip install submodules/simple-knn
pip install submodules/raytracing

# install other denpendencies
pip install -r requirements.txt

Dateset

We mainly test our method on Shiny Blender Synthetic, Shiny Blender Real, Glossy Synthetic and NeRF Synthetic dataset. Please run the script nero2blender.py to convert the format of the Glossy Synthetic dataset.

Running

We provide the script to test our code on each scene of datasets. Just run:

sh train.sh

You may need to modify the path in train.sh

Command Line Arguments for train.py

--iteration

The number of total iteration for training.

--lambda_normal_smooth

The strength of normal smooth loss.

Evaluation

python eval.py --white_background --save_images --model_path output/NAME_OF_THE_SCENE

You will get PSNR/SSIM/LPIPS/FPS results.

Acknowledgement

This work is built on many amazing research works:

• 3DGS-DR
• 2DGS
• Raytracing

📜 BibTeX

@inproceedings{yao2025refGS,
  title={Reflective Gaussian Splatting},
  author={Yao, Yuxuan and Zeng, Zixuan and Gu, Chun and Zhu, Xiatian and Zhang, Li},
  booktitle={ICLR},
  year={2025},
}