Concept-Aware Learning for Weakly Supervised Video Anomaly Detection

doi:10.1016/j.patcog.2026.113853

Concept-Aware Learning for Weakly Supervised Video Anomaly Detection

2026-04-27

0

PRE

AI

DOI：10.1016/j.patcog.2026.113853

Original

Original request for help

Save

Abstract

En 中文

• Our method enables explicit modeling of anomalous components and improves generalization. • We propose the concepts of non-learnable text and learnable prompt to leverage anomalous priors. • To effectively capture temporal dependency features, we improve a temporal encoder.

Keywords:

anomaly detection

weakly supervised learning

temporal dependency

prompt learning

concept-aware learning

Concept-Aware Learning for Weakly Supervised Video Anomaly Detection

Abstract

Journal

Organization