GS-I³：: Gaussian Splatting for Surface Reconstruction from Illumination-Inconsistent Images

Tengfei Wang, Hongmao Hou, Zhaoning Zhang, Yiwei Xu, Zongqian Zhan and Xin Wang*.

Project Page | arXiv

Data

Dark in the Gaussian dataset can be downloaded from DK-Gaussian .

The Modified DTU dataset can be downloaded from Modified DTU .

Installation

The repository contains submodules, thus please check it out with

# SSH
git@github.com:TFwang-9527/GS-3I.git
cd GS-3I

conda create -n GS-3I python=3.8
conda activate GS-3I

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 #replace your cuda version
pip install -r requirements.txt
pip install submodules/diff-plane-rasterization
pip install submodules/simple-knn

Preprocessing

First, we need to obtain the pretrain normal map from StableNormal or other methods. Then, place the normals folder and the images folder under the same directory.

Training

python train.py -s data_path -m out_path --max_abs_split_points 0 --opacity_cull_threshold 0.05

Rendering and Extract Mesh

python render.py -m out_path --max_depth 10.0 --voxel_size 0.01

Acknowledgements

This project is built upon 3DGS, PGSR, StableNormal, and Gaussian in the Dark. respectively. We thank all the authors for their great work and repos.