Analýza zornic v obrazech

Abstract

The aim of this thesis was to get familiar with pupils detection methods and then implement a pupil detector with usage of available libraries. There are several solutions proposed in this thesis and libraries used in the implementation are OpenCV and dlib library. A combination of evolutionary algorithms with pupils detection methods is used in the solution. Presented methods use different kinds of detection, some of them are based on image analysis primary algorithms such as edge detection, the other solutions use machine learning methods. Precision and speed of the proposed solutions with usage of different algorithms parameters settings is verified on renowned data sets.

Description

Subject(s)

pupils detection, evolutionary algorithms, edges detection, HOG descriptor, SVM classifier, convolutional neural network

Citation