Adversarial composite prediction of normal video dynamics for anomaly detection

Gang Li, Ping He, Huibin Li, Fan Zhang

Abstract: Highlights•We propose a novel adversarial composite prediction framework for unsupervised video anomaly detection, namely ACP-VAD. It consists of an ACP module and a FMML module.•ACP module can reasonably model normal video dynamics through adversarial learning and bidirectional composite prediction strategy.•FMML module can learn a more compact representation of normal video dynamics and improve the detection sensibility of motion anomalies.