🌐Project Page | 🖨️ArXiv | 📰Paper
# install requirements
conda env create -n enviroment.yaml
conda activate SK_GS
# build extension
cd <project root>
cd extenstion/_C
mkdir build
cd build
cmake ..
make -j
- Download D-NeRF dataset. Unzip the downloaded data tor prooject root data dir in order to train.
- Download WIM dataset and Unzip to dir.
- Prepare ZJU Mocap dataset as watch-it-move
- Dataset structure
<project root>
├── data
│ ├── DNeRF
│ │ ├── mutant
│ │ ├── standup
│ │ ├── ...
│ ├── WIM
│ │ ├── atlas
│ │ ├── baxter
│ │ ├── ...
│ ├── zju
│ │ ├── 313
│ │ ├── ...
python train.py -c exps/d_nerf.yaml --scene hook
python test.py -c exps/d_enrf.yaml --scene hook --load results/DNeRF/last.pth
python gui.py -c exps/d_enrf.yaml --scene hook --load results/DNeRF/last.pth
Thanks to the authors of 3D Gaussians, Deformable-3D-Gaussians and SC-GS for their excellent code.
@InProceedings{SK-GS,
title = {Template-free Articulated Gaussian Splatting for Real-time Reposable Dynamic View Synthesis},
author = {Wan, Diwen and Wang, Yuxiang and Lu, Ruijie and Zeng, Gang},
booktitle = {NeurIPS},
year = {2024},
}
@InProceedings{SP-GS,
title = {Superpoint Gaussian Splatting for Real-Time High-Fidelity Dynamic Scene Reconstruction},
author = {Wan, Diwen and Lu, Ruijie and Zeng, Gang},
booktitle = {Proceedings of the 41st International Conference on Machine Learning},
pages = {49957--49972},
year = {2024},
}