1 paper
Yuandu Lai, Yahong Han, Yaowei Wang
Recent efforts towards video anomaly detection (VAD) try to learn a deep autoencoder to describe normal event patterns with small reconstruction errors. The video inputs with large…