TASAM: Terrain-and-Aware Segment Anything Model for Temporal-Scale Remote Sensing Segmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Tianyang, Xiao, Xi, Chen, Gaofei, Chi, Hanzhang, Zhang, Qi, Cheng, Guo, Ji, Yingrui
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912594610094080
author Wang, Tianyang
Xiao, Xi
Chen, Gaofei
Chi, Hanzhang
Zhang, Qi
Cheng, Guo
Ji, Yingrui
author_facet Wang, Tianyang
Xiao, Xi
Chen, Gaofei
Chi, Hanzhang
Zhang, Qi
Cheng, Guo
Ji, Yingrui
contents Segment Anything Model (SAM) has demonstrated impressive zero-shot segmentation capabilities across natural image domains, but it struggles to generalize to the unique challenges of remote sensing data, such as complex terrain, multi-scale objects, and temporal dynamics. In this paper, we introduce TASAM, a terrain and temporally-aware extension of SAM designed specifically for high-resolution remote sensing image segmentation. TASAM integrates three lightweight yet effective modules: a terrain-aware adapter that injects elevation priors, a temporal prompt generator that captures land-cover changes over time, and a multi-scale fusion strategy that enhances fine-grained object delineation. Without retraining the SAM backbone, our approach achieves substantial performance gains across three remote sensing benchmarks-LoveDA, iSAID, and WHU-CD-outperforming both zero-shot SAM and task-specific models with minimal computational overhead. Our results highlight the value of domain-adaptive augmentation for foundation models and offer a scalable path toward more robust geospatial segmentation.
format Preprint
id arxiv_https___arxiv_org_abs_2509_15795
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TASAM: Terrain-and-Aware Segment Anything Model for Temporal-Scale Remote Sensing Segmentation
Wang, Tianyang
Xiao, Xi
Chen, Gaofei
Chi, Hanzhang
Zhang, Qi
Cheng, Guo
Ji, Yingrui
Computer Vision and Pattern Recognition
Segment Anything Model (SAM) has demonstrated impressive zero-shot segmentation capabilities across natural image domains, but it struggles to generalize to the unique challenges of remote sensing data, such as complex terrain, multi-scale objects, and temporal dynamics. In this paper, we introduce TASAM, a terrain and temporally-aware extension of SAM designed specifically for high-resolution remote sensing image segmentation. TASAM integrates three lightweight yet effective modules: a terrain-aware adapter that injects elevation priors, a temporal prompt generator that captures land-cover changes over time, and a multi-scale fusion strategy that enhances fine-grained object delineation. Without retraining the SAM backbone, our approach achieves substantial performance gains across three remote sensing benchmarks-LoveDA, iSAID, and WHU-CD-outperforming both zero-shot SAM and task-specific models with minimal computational overhead. Our results highlight the value of domain-adaptive augmentation for foundation models and offer a scalable path toward more robust geospatial segmentation.
title TASAM: Terrain-and-Aware Segment Anything Model for Temporal-Scale Remote Sensing Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.15795