TASAM: Terrain-and-Aware Segment Anything Model for Temporal-Scale Remote Sensing Segmentation
This addresses the need for robust geospatial segmentation in remote sensing, offering a domain-adaptive augmentation for foundation models, but it is incremental as it builds on SAM with lightweight modules.
The paper tackled the problem of Segment Anything Model (SAM) struggling with remote sensing data due to terrain, multi-scale objects, and temporal dynamics, by introducing TASAM, which achieved substantial performance gains across three benchmarks (LoveDA, iSAID, WHU-CD) with minimal computational overhead.
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.