Manual crater detection in planetary science is laborious and time-consuming. This paper introduces a universal crater detection scheme using Meta AI's Segment Anything Model (SAM). SAM's zero-shot generalization capabilities allow crater identification across diverse datasets (images, DEMs) and celestial bodies (Moon, Mars) without retraining. Shape indexes filter segmentation masks, retaining only crater-like features, which are then fitted with ellipses to determine crater location and size. The method demonstrates effectiveness across various datasets, highlighting SAM's potential for planetary science.
