GeoSeg: Training-Free Reasoning-Driven Segmentation in Remote Sensing Imagery

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jiang, Lifan, Pei, Yuhang, Wu, oxi, Zhao, Yan, Wu, Tianrun, Yu, Shulong, Zhang, Lihui, Cai, Deng
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917313814462464
author Jiang, Lifan
Pei, Yuhang
Wu, oxi
Zhao, Yan
Wu, Tianrun
Yu, Shulong
Zhang, Lihui
Cai, Deng
author_facet Jiang, Lifan
Pei, Yuhang
Wu, oxi
Zhao, Yan
Wu, Tianrun
Yu, Shulong
Zhang, Lihui
Cai, Deng
contents Recent advances in MLLMs are reframing segmentation from fixed-category prediction to instruction-grounded localization. While reasoning based segmentation has progressed rapidly in natural scenes, remote sensing lacks a generalizable solution due to the prohibitive cost of reasoning-oriented data and domain-specific challenges like overhead viewpoints. We present GeoSeg, a zero-shot, training-free framework that bypasses the supervision bottleneck for reasoning-driven remote sensing segmentation. GeoSeg couples MLLM reasoning with precise localization via: (i) bias-aware coordinate refinement to correct systematic grounding shifts and (ii) a dual-route prompting mechanism to fuse semantic intent with fine-grained spatial cues. We also introduce GeoSeg-Bench, a diagnostic benchmark of 810 image--query pairs with hierarchical difficulty levels. Experiments show that GeoSeg consistently outperforms all baselines, with extensive ablations confirming the effectiveness and necessity of each component.
format Preprint
id arxiv_https___arxiv_org_abs_2603_03983
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle GeoSeg: Training-Free Reasoning-Driven Segmentation in Remote Sensing Imagery
Jiang, Lifan
Pei, Yuhang
Wu, oxi
Zhao, Yan
Wu, Tianrun
Yu, Shulong
Zhang, Lihui
Cai, Deng
Computer Vision and Pattern Recognition
Artificial Intelligence
Recent advances in MLLMs are reframing segmentation from fixed-category prediction to instruction-grounded localization. While reasoning based segmentation has progressed rapidly in natural scenes, remote sensing lacks a generalizable solution due to the prohibitive cost of reasoning-oriented data and domain-specific challenges like overhead viewpoints. We present GeoSeg, a zero-shot, training-free framework that bypasses the supervision bottleneck for reasoning-driven remote sensing segmentation. GeoSeg couples MLLM reasoning with precise localization via: (i) bias-aware coordinate refinement to correct systematic grounding shifts and (ii) a dual-route prompting mechanism to fuse semantic intent with fine-grained spatial cues. We also introduce GeoSeg-Bench, a diagnostic benchmark of 810 image--query pairs with hierarchical difficulty levels. Experiments show that GeoSeg consistently outperforms all baselines, with extensive ablations confirming the effectiveness and necessity of each component.
title GeoSeg: Training-Free Reasoning-Driven Segmentation in Remote Sensing Imagery
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2603.03983