TL;DR本文提出了一种基于 SAM 模型和 Prompt 学习的自动遥感图像实例分割方法(RSPrompter),并将其与其他常用的方法进行比较,结果证明该方法有效。
Abstract
Leveraging vast training data (SA-1B), the foundation Segment Anything Model (SAM) proposed by Meta AI Research exhibits remarkable generalization and zero-shot capabilities. Nonetheless, as a category-agnostic instance segmentation method, SAM heavily depends on prior manual guidance