Guangchi Fang 方广驰

Interested in 3D vision, particularly for 3D data processing inlcuding point cloud and neural representation.

Email / Github / Google Scholar

Recent Work
PontTuset

Mini-Splatting: Representing Scenes with a Constrained Number of Gaussians
G. Fang and B. Wang
ECCV 2024
ArXiv / Code

Point cloud analysis in the context of Gaussian Splatting. Reorganizing the spatial distribution of Gaussians to construct an efficient scene representation.


3DGS Mini-Splatting-D Mini-Splatting

Check out the teaser for details of spatial distribution.


Quantitative evaluation. All experimental settings are consistent with the original 3DGS implementation.


Publications
PontTuset

ACRF: Compressing Explicit Neural Radiance Fields via Attribute Compression
G. Fang, Q. Hu, L. Wang, Y. Guo
ICLR 2024
Paper / Code

Compression of neural radiance fields via point cloud processing and 3D compression techniques.
PontTuset

3DAC: Learning Attribute Compression for Point Clouds
G. Fang, Q. Hu, H. Wang, Y. Xu, Y. Guo
CVPR 2022
Paper / Code / Poster

Point cloud attribute compression leveraging deep entropy coding.
PontTuset

3DPointCaps++: Learning 3D Representations with Capsule Networks
Y. Zhao*, G. Fang*, Y. Guo, L. Guibas, F. Tombari, T. Birdal
IJCV 2022 (* indicates equal contribution)
Paper / Code

Rigid and non-rigid point cloud representation learning based on capsule.

PontTuset

SQN: Weakly-Supervised Semantic Segmentation of Large-Scale 3D Point Clouds
Q. Hu, B. Yang, G. Fang, A. Leonardis, Y. Guo, N. Trigoni, A. Markham
ECCV 2022
ArXiv / Code

Weakly supervised point cloud segmentation through point query.

Thanks.