Assessment of the Segment Anything Model's Applicability to Coastal Debris Segmentation:A Comparative Analysis of Prompt Input Methods

초록

Coastal debris poses serious threats to marine ecosystems and fisheries, increasing the demandfor effective detection technologies. However, the irregular shapes, low color contrast, and complexbackgrounds of such debris limit the generalization capability of supervised segmentation models. Toaddress this limitation, this study evaluates the applicability of the Segment Anything Model (SAM),which performs object segmentation based solely on prompt inputs without requiring additional training. The segmentation performance of SAM was quantitatively assessed under various conditions by comparingGround Truth (GT)-based prompts with automatically generated prompts. Although automatic promptsgenerally yielded lower performance than GT-based ones, relatively high accuracy was observed forobjects with distinct color contrast or simple backgrounds. In particular, the automatic point-basedprompt achieved performance comparable to GT for certain object categories. These results suggest that,with well-designed prompts, SAM can produce a certain level of segmentation performance even incomplex coastal environments. This implies that SAM may be useful for preliminary object detection inpreviously unseen regions where labeled data are unavailable. Furthermore, optimizing prompt designin combination with post-processing strategies could improve its practical applicability in coastal areaslacking sufficient training data.

키워드