Analýza anomálních stavů řidiče pomocí obrazů

Abstract

This master thesis explores different approaches to the analysis of anomalous driver states using images. Driver monitoring can prevent safety risks associated with driver inattention and health problems. The experimental part of the thesis focuses on anomaly detection using machine learning and it is divided into two main parts. In the first part, the use of supervised learning methods is examined and the next part of the work is dedicated to unsupervised learning. The proposed solutions are assessed using several evaluation metrics and also in terms of computational complexity. Based on the performance of the models, a program is developed to analyze driver behavior using the appropriate technologies.