MirrorGS

An unofficial implementation of Mirror3DGS: Incorporating Mirror Reflections into 3D Gaussian Splatting.

output.mp4

output.mp4

Implementation Notes

You can find detailed explainations from this blog.

Install

git clone --recursive https://github.com/TingtingLiao/MirrorGS.git 
cd MirrorGS
conda create -n mirrorgs python=3.10 
conda activate mirrorgs 

# torch2.3.0+cu12.1 
pip install torch torchvision torchaudio
 
# requirements
pip install -r requirements.txt

# xformers  
pip install -U xformers --index-url https://download.pytorch.org/whl/cu121
 
pip install -e submodules/diff-surfel-rasterization 
pip install -e submodules/simple-knn 

Data

We provide an example. Please download it from here.

For customized data, the following steps are required:

• mirror segmention
• remove the mirror area (the right image), then run colmap

colmap automatic_reconstructor --workspace_path data/{}  --image_path data/{}/images --camera_model SIMPLE_PINHOLE  

Usage

python train.py -s data/{}   
  
# validation 
python render.py -m ./output/{your_path} -s data/{} --skip_mesh  

Acknowledgement

Special thanks to the projects and their contributors:

• Mirror3DGS
• MirrorGaussian
• 2DGS
• pyRANSAC-3D