- Integrate depth, alpha, and segmentation rendering
- Maintain rendered image names
- Log training with wandb
- Depth, alpha, and segmentation rendering
- Video rendering
# SSH
git clone [email protected]:weijielyu/gaussian-splatting.git --recursive
or
# HTTPS
git clone https://github.com/weijielyu/gaussian-splatting --recursive
conda create -n gs python=3.8 -y
conda activate gs
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch
pip install -r requirement.txt
pip install submodules/simple-knn
- For vanilla Gaussian Splatting rendering:
pip install submodules/diff-gaussian-rasterization
- For depth, alpha rendering with confidence:
pip install submodules/diff-gaussian-rasterization-confidence
- For segmentation rendering:
pip install submodules/diff-gaussian-rasterization-gaga
python convert.py -s <location> --skip_matching [--resize] #If not resizing, ImageMagick is not needed
python train.py -s <path to COLMAP or NeRF Synthetic dataset>
python render.py -m <path to pre-trained model> -s <path to COLMAP dataset>
python metrics.py -m <path to pre-trained model>
This repository is modified based on the following repositories. Thanks for their wonderful implementations!
https://github.com/graphdeco-inria/gaussian-splatting
https://github.com/VITA-Group/FSGS
https://github.com/lkeab/gaussian-grouping