Rozpoznání obličejů a jejich sledování na procesorech i.MX

Abstract

This bachelor thesis deals with detecting and recognizing human faces. First of all, known and effective methods are described, which include older breakthrough methods like Viola-Jones for detection or Eigenfaces for recognition, but also modern methods like neural networks. After that, methods are implemented using OpenCV libraries and tested to find optimal parameters on FDDB database for face detection and on AT&T database for face recognition. With these parameters, the methods are tested using the Raspberry Pi v2 camera module in a real environment. All testing was carried out on device Raspberry Pi 3 model B with i.MX processor.