GeoSeg: Training-Free Reasoning-Driven Segmentation in Remote Sensing Imagery
Lifan Jiang, Yuhang Pei, Boxi Wu, Yan Zhao, Tianrun Wu, Shulong Yu, Lihui Zhang, Deng Cai · 2026
Read GeoSeg: Training-Free Reasoning-Driven Segmentation in Remote Sensing Imagery by Lifan Jiang, Yuhang Pei, Boxi Wu, Yan Zhao, Tianrun Wu, Shulong Yu, Lihui Zhang, Deng Cai (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-2603-03983.txt.
Description
Lifan Jiang and others present GeoSeg, a training-free method for thought-driven segmentation in remote sensing imagery. Reasoning-based segmentation in remote sensing still lacks a broadly applicable answer because data are costly and viewpoints are overhead. GeoSeg couples model thought with precise localization through bias-aware coordinate refinement correcting systematic grounding shifts and a dual-route prompting process fusing meaning intent with fine-grained spatial cues. The team also presents GeoSeg-Bench, a diagnostic set of 810 image-query pairs with graded difficulty.
Catalog record
| CREATOR | Lifan Jiang, Yuhang Pei, Boxi Wu, Yan Zhao, Tianrun Wu, Shulong Yu, Lihui Zhang, Deng Cai |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52295 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.03983