Geometry-Guided Self-Supervision for Ultra-Fine-Grained Recognition with Limited Data
Shijie Wang 1,2 , Yadan Luo 2 , Zijian Wang 2 , Haojie Li 1,∗ , Zi Huang 2 , Mahsa Baktashmotlagh 2 1 Shandong University of Science and Technology, China 2 The University of Queensland, Australia · 2026
Read Geometry-Guided Self-Supervision for Ultra-Fine-Grained Recognition with Limited Data by Shijie Wang 1,2 , Yadan Luo 2 , Zijian Wang 2 , Haojie Li 1,∗ , Zi Huang 2 , Mahsa Baktashmotlagh 2 1 Shandong University of Science and Technology, China 2 The University of Queensland, Australia (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-2604-19345.txt.
Description
Wang and colleagues study the inbuilt shape features of highly similar objects for ultra-fine-grained sorting with limited data. Their Geometric Attribute Exploration Network makes shape attributes as fresh recognition cues, drawn from details that match an object's patterns, such as vein structures in soybean leaves. Each class shows distinct shape signs that stay strong even among near-twin objects. The tool finds those attributes by first boosting shape-key signals, a factor recent work has largely missed.
Catalog record
| CREATOR | Shijie Wang 1,2 , Yadan Luo 2 , Zijian Wang 2 , Haojie Li 1,∗ , Zi Huang 2 , Mahsa Baktashmotlagh 2 1 Shandong University of Science and Technology, China 2 The University of Queensland, Australia |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44189 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.19345