Weakly Supervised From survey PDF — verify links
CMSIL
CMSIL — Vision-Language Adaptation
ICME · 2024Human-reviewed summary status
Survey summary
Two-branch framework pairing a CLIP-driven vision-language branch, which generates pseudo anomaly cues from visual-concept priors, with multi-scale instance learning that models temporal correlation between adjacent snippets to reduce the false alarms typical of standard MIL.
Taxonomy placement
Weakly Supervised
Weakly SupervisedVision-Language Adaptation
Method characteristics
Weakly SupervisedCLIP-drivenMulti-scale instance learning