Abstract: This letter provides insights on the effectiveness of the zero-shot, prompt-based Segment Anything Model (SAM) and its updated versions, SAM 2 and SAM 2.1, along with the nonpromptable ...