SAMBLE: Shape-Specific Point Cloud Sampling for an Optimal Trade-Off Between Local Detail and Global Uniformity

Chengzhi Wu, Yuxin Wan, Hao Fu, Julius Pfrommer, Zeyun Zhong, Junwei Zheng, Jiaming Zhang, Jürgen Beyerer · CVPR 2025 · 2025

DOI arXiv

Status

To read

Topics

sampling

Feeds

First paper · height sampling, P1 · ICCV 2027

What they did

Proposes SAMBLE, a learned point sampling method that adapts its strategy to each shape, balancing local detail such as edges against global uniformity, and trains jointly with the downstream task.

My note

Learns sampling strategies that balance detailed regions with global coverage. An important learned-sampling baseline.

BibTeX

@inproceedings{wu2025-samble,
  title = {{SAMBLE: Shape-Specific Point Cloud Sampling for an Optimal Trade-Off Between Local Detail and Global Uniformity}},
  author = {Chengzhi Wu and Yuxin Wan and Hao Fu and Julius Pfrommer and Zeyun Zhong and Junwei Zheng and Jiaming Zhang and Jürgen Beyerer},
  booktitle = {CVPR 2025},
  year = {2025},
  doi = {10.1109/CVPR52734.2025.00133},
  eprint = {2504.19581},
  archivePrefix = {arXiv}
}