ClimaOoD: Improving Anomaly Segmentation via Physically Realistic Synthetic Data
Yuxing Liu Yong Liu · 2025
Read ClimaOoD: Improving Anomaly Segmentation via Physically Realistic Synthetic Data by Yuxing Liu Yong Liu (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-2512-02686.txt.
Description
Yuxing Liu and Yong Liu improve anomaly segmentation, the task of finding odd objects on the road. Self driving systems must detect things outside their known classes, since the open world always holds surprises. Real anomaly data is scarce and narrow, which limits how well models generalize.
ClimaOoD creates physically realistic synthetic anomaly data to fill that gap. Training on varied, plausible oddities teaches models to flag the truly unknown. Better anomaly segmentation is a safety issue: a system that notices what it does not know can slow down or hand over control in time. Fake odd cases must look real, or the model learns to spot the fake, not the road risk ahead.
Catalog record
| CREATOR | Yuxing Liu Yong Liu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-02686