Voxel-GS: Quantized Scaffold Gaussian Splatting Compression with Run-Length Coding DCC26

Dataset

voxelGS/data
├── DeepBlending 
    ├── drjohnson
    └── playroom
├── MipNerf360 
    ├── bicycle
    ├── bonsai
    ├── counter
    ├── flowers
    ├── garden
    ├── kitchen
    ├── room
    ├── stump
    └── treehill
├── Nerf_Synthetic 
    ├── chair
    ├── drums
    ├── ficus
    ├── hotdog
    ├── lego
    ├── materials
    ├── mic
    └── ship
└── T2T 
    ├── train
    └── truck

Environment

cuda 11.7+python3.8

conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia
pip install omegaconf loguru open3d opencv-python plyfile tensorboard termcolor torch_scatter jaxtyping einops lpips
pip install models/submodules/* # from Scaffold-GS

Train

1. Run python Train.py. After training, stat_log will generate statistical information and save the results in the output folder.
2. Note: Due to the randomness of the training process, there may be slight differences in the results.
3. Alternatively, you can download and unzip Results/BIN.zip to directly access the compressed bin and decompress it using the command: python Coder.py -bin Results/BIN/DeepBlending/playroom/gsbin -eval

Encoding, decoding, evaluation, show from separated GS PLY

1. Provide -ply -encode for encoding
   python Coder.py -ply Results/BIN/Nerf_Synthetic/hotdog/point_cloud/point_cloud_30000.quantized.ply -encode

2. Provide -bin for decoding
   python Coder.py -bin Results/BIN/Nerf_Synthetic/hotdog/gsbin

3. Add -show -eval for visualization and evaluation
   python Coder.py -bin Results/BIN/Nerf_Synthetic/hotdog/gsbin -show -eval