Rozpoznání směru pohledu řidiče s využitím kamery ve vozidle

Abstract

This bachelor thesis deals with the estimation of driver's gaze using deep learning technology. The gaze monitoring can reduce the risks of driver's inattention and prevent potential hazards. The paper is focused on existing methods for eye gaze estimation, the implementations of which are freely available. The proposed methods are presented and tested on custom video recordings taken while driving a vehicle. Based on this testing, success of the methods is evaluated, which depends not only on the accuracy of the estimated gaze vectors, but also the time required, which is important for real-time monitoring.