Reconstruction-Based Video Anomaly Detection Approaches

In this chapter, we present a number of techniques that make use of a set of deep learning neural network approaches, named autoencoder, for representation-based anomaly detection that continues the work presented in Chap. 9 by using information extracted from observed normal data through deep neural networks to provide more effective and efficient anomaly detection solutions. We will look at a number of deep neural network architectures principally designed for the representation of the newly coming video images.