Detekce obličejových bodů v prostředí automobilu

Abstract

The proposed bachelor thesis is focused on detections, respectively predictions of facial landmarks in video sequences by using the technics of machine learning. The thesis is dedicated to existing methods of facial landmark detections whose implementations are freely available on the internet. The success rate of proposed methods which have been tested on the custom video sequences taken in the realworld conditions of traffic are presented. With the term of success rate is meant the accuracy of predicted facial landmarks against to manually annotated corresponding landmarks.