Welcome! The official implementation of the paper "ContextGS: Compact 3D Gaussian Splatting with Anchor Level Context Model" will be released here soon!

Yufei Wang, Zhihao Li, Lanqing Guo, Wenhan Yang, Alex C. Kot, Bihan Wen

Our method, ContextGS, first proposes to reduce the spatial redundancy among anchors using an autoregressive model.

We divide anchors into levels as shown in Fig.(b) and the anchors from coarser levels are used to predict anchors in finer levels, i.e., red anchors predict blue anchors then red and blue anchors together predict black anchors. Fig.(c) verifies the spatial redundancy by calculating the cosine similarity between anchors in level $0$ and their context anchors in levels $1$ and $2$. Fig.(d) displays the bit savings using the proposed anchor-level context model evaluated on our entropy coding based strong baseline built on Scaffold-GS.

Compared with Scaffold-GS, we achieve better rendering qualities, faster rendering speed, and great size reduction of up to $15$ times averaged over all datasets we used.

The code will be released soon.

Please cite our paper if you find our work useful. Thanks!

@article{wang2024contextgs,
  title={ContextGS: Compact 3D Gaussian Splatting with Anchor Level Context Model},
  author={Wang, Yufei and Li, Zhihao and Guo, Lanqing and Yang, Wenhan and Kot, Alex C and Wen, Bihan},
  journal={arXiv preprint arXiv:2405.20721},
  year={2024}
}

If you have any questions, please feel free to contact me via [email protected].