High-quality one-shot interactive segmentation for remote sensing images via hybrid adapter-enhanced foundation models

Zhili Zhang, Xiangyun Hu, Yue Yang, Bingnan Yang, Kai Deng, Hengming Dai, Mi Zhang

Abstract: Highlights•A one-shot interactive segmentation method enables high-quality remote sensing results.•A plug-and-play hybrid adapter enhances foundation models for interactive segmentation.•Diverse prompts (points, boxes, scribbles) are encoded uniformly for robust performance.•The method improves annotation efficiency for remote sensing image labeling.