DOSE3 : Diffusion-based Out-of-distribution detection on ๐ โ ๐ผ โ ( 3 ) \mathbb{SE}(3) trajectories
Hongzhe Cheng, Tianyou Zheng *, Tianyi Zhang, Matthew Johnson-Roberson, Weiming Zhi ยท 2025
Read DOSE3 : Diffusion-based Out-of-distribution detection on ๐ โ ๐ผ โ ( 3 ) \mathbb{SE}(3) trajectories by Hongzhe Cheng, Tianyou Zheng *, Tianyi Zhang, Matthew Johnson-Roberson, Weiming Zhi (2025) 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-2502-16725.txt.
Description
Hongzhe Cheng and four co-authors extend a safety tool for robots and vision systems, in a 2025 paper. Out-of-distribution detection flags inputs that differ from what a system was trained on, before they cause errors. Diffusion models have proved good at this for images, but robot work involves motion through space, not flat pictures. A pose sequence lives in a space that mixes position and rotation, known as SE(3).
Their framework, DOSE3, adapts diffusion scoring to such pose trajectories. Tests on several benchmark datasets show it can spot abnormal motion paths. Reliable oddity detection matters for self-driving tools and warehouse robots, where a strange movement may signal a fault or a hazard that the system should not ignore.
Catalog record
| CREATOR | Hongzhe Cheng, Tianyou Zheng *, Tianyi Zhang, Matthew Johnson-Roberson, Weiming Zhi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-16725