Object-Centric Industrial Video Anomaly Detection with Local-Global Representation Learning
A novel framework centered on object-centric video anomaly detection, heavily augmented by a local-global representation learning mechanism, suggesting that integrating structured object interactions into representation learning provides a highly scalable and robust solution for real-world industrial monitoring.