Rozpoznávání anomálií v lidském chování v sekvencích obrazů získaných kamerou

Abstract

This work is about detecting anomalies in driver behavior. Firstly, I discuss the general description of the problem and its challenges. Then, I briefly address the functionality of neural networks and subsequently describe my own solution using a neural network. I also describe an alternative way in which the neural network could be replaced by a Gaussian mixture model. Finally, I demonstrate how it is possible to detect anomalies by an interesting combination of these two approaches.