PointCSP: Cross-Sample Semantic Propagation and Stability Preservation in Self-Supervised Point Cloud Learning
Xinxing Yu, Ajian Liu, Sunyuan Qiang, Hui Ma, Liying Yang, Yuzhong Wang, Zhi Rao, Yanyan Liang · 2026
Read PointCSP: Cross-Sample Semantic Propagation and Stability Preservation in Self-Supervised Point Cloud Learning by Xinxing Yu, Ajian Liu, Sunyuan Qiang, Hui Ma, Liying Yang, Yuzhong Wang, Zhi Rao, Yanyan Liang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-01759.txt.
Description
Xinxing Yu and seven co-authors teach 3D scene models by letting point clouds learn from each other. Self-supervised training on single scenes, one at a time, keeps each scene's meaning in its own dialect, which blocks the shared semantic space that transfer needs.
Their PointCSP framework serializes the samples in a batch into one continuous stream and runs a state-space model over it, so semantic state propagates across samples and meanings line up globally. That batch trick cannot survive to single-scene testing. A second stage distills the pretrained semantics back into a per-scene form, using mixed inputs and a feature alignment constraint to keep the transfer stable. Across several benchmark sets the method beats the state of the art on score and on semantic consistency alike. The design targets scene-level 3D vision, where labels are dear and consistency across places is the prize.
Catalog record
| CREATOR | Xinxing Yu, Ajian Liu, Sunyuan Qiang, Hui Ma, Liying Yang, Yuzhong Wang, Zhi Rao, Yanyan Liang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.01759