Weakly Supervised From survey PDF — verify links
TEVAD
TEVAD — Caption-Guided Semantics
CVPR · 2023Human-reviewed summary status
Survey summary
TEVAD (Text-Empowered VAD) generates dense captions for video snippets and fuses text and visual features for anomaly detection, improving accuracy and robustness on ShanghaiTech, UCF-Crime, XD-Violence, and UCSD-Pedestrians while using word-level caption contributions to explain predictions.
Taxonomy placement
Weakly Supervised
Weakly SupervisedCaption-Guided Semantics
Method characteristics
Weakly SupervisedDense captionsSentence embeddingsInterpretability