Mohamed Mostafa Fouad

HRM-Crime: A Hierarchical Relationship Model for Weakly Supervised Crime Anomaly Detection in Surveillance Video

Detecting anomalous behaviour in surveillance video is a high-impact application of machine learning, but supervision is available only at the video level: a set of temporal segments carries a single anomaly or normal label, with no per-segment ground truth. This is the Multiple Instance Learning setting. On the UCF-Cr...