weijielyu/gaussian-splatting

Original reference implementation of "3D Gaussian Splatting for Real-Time Radiance Field Rendering"

My Modified Version of 3DGS

Todos

• Integrate depth, alpha, and segmentation rendering

Added Features

• Maintain rendered image names
• Log training with wandb
• Depth, alpha, and segmentation rendering
• Video rendering

Commands

Cloning the Repository

# SSH
git clone [email protected]:weijielyu/gaussian-splatting.git --recursive

or

# HTTPS
git clone https://github.com/weijielyu/gaussian-splatting --recursive

Environment

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

Colmap

python convert.py -s <location> --skip_matching [--resize] #If not resizing, ImageMagick is not needed

Training

python train.py -s <path to COLMAP or NeRF Synthetic dataset>

Evaluation

python render.py -m <path to pre-trained model> -s <path to COLMAP dataset>
python metrics.py -m <path to pre-trained model>

Acknowledge

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