Noise-Robust Tiny Object Localization with Flows Journal: Pattern Recognition
Huixin Sun 1 , Linlin Yang 2 , Ronyu Chen 3 , Kerui Gu 3 , Baochang Zhang 1 , Angela Yao 3 , Xianbin Cao 1 1 Beihang University, China 2 Communication University of China, China 3 National University of Singapore, Singapore · 2026
Read Noise-Robust Tiny Object Localization with Flows Journal: Pattern Recognition by Huixin Sun 1 , Linlin Yang 2 , Ronyu Chen 3 , Kerui Gu 3 , Baochang Zhang 1 , Angela Yao 3 , Xianbin Cao 1 1 Beihang University, China 2 Communication University of China, China 3 National University of Singapore, Singapore (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-2601-00617.txt.
Description
Huixin Sun and six co-authors address a weak spot of AI vision in a 2026 paper: finding tiny objects. Detectors do well on objects of normal size, but lag on tiny ones. The team show that tiny objects are hurt most by noisy labels. A strict aim on exact place can make a model fit that noise too well. Their method, named TOLF, models the error in a flexible way.
TOLF uses a tool that can bend to fit odd error shapes, not just bell curves. It also weighs each training case by how unsure the model is. Cases with high doubt, which are prone to noise, count for less. Tests on three data sets back the approach. On one standard set of tiny-object photos, TOLF raised a strong base model's score. The work points to noise care as a key step for tiny-object vision.
Catalog record
| CREATOR | Huixin Sun 1 , Linlin Yang 2 , Ronyu Chen 3 , Kerui Gu 3 , Baochang Zhang 1 , Angela Yao 3 , Xianbin Cao 1 1 Beihang University, China 2 Communication University of China, China 3 National University of Singapore, Singapore |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-00617