SAM Struggles in Concealed Scenes -- Empirical Study on Segment Anything
This study identifies a critical limitation of SAM for applications in domains like wildlife monitoring, industrial inspection, and medical imaging, though it is incremental as it focuses on empirical testing rather than proposing a new method.
The researchers evaluated the Segment Anything Model (SAM) on three concealed scenes—camouflaged animals, industrial defects, and medical lesions—and found that SAM performs poorly in these unprompted settings.
Segmenting anything is a ground-breaking step toward artificial general intelligence, and the Segment Anything Model (SAM) greatly fosters the foundation models for computer vision. We could not be more excited to probe the performance traits of SAM. In particular, exploring situations in which SAM does not perform well is interesting. In this report, we choose three concealed scenes, i.e., camouflaged animals, industrial defects, and medical lesions, to evaluate SAM under unprompted settings. Our main observation is that SAM looks unskilled in concealed scenes.