Ανάπτυξη συστήματος ταυτοποίησης προσώπων σε ροές βίντεο με χρήση Συνελικτικών Νευρωνικών Δικτύων
Development of a face identification system for video streams using Convolutional Neural Networks
View/ Open
Abstract
The subject of the present Master Thesis is the development of a Face Recognition application for video files, implementing Convolutional Neural Networks through the use of the OpenFace library. The goal of this Thesis is to present an implementation of open source technologies that could provide comparable results to other state-of-the-art private face recognition systems, while being well-suited for mobile scenarios. The application that is presented, provides the user with a fully realized framework for the search, preprocessing, training and subsequent classification of faces, without the need for external tools. For the training stage of the classifier the Linear SVM algorithm was chosen, because it provided the best results compared to other algorithms while providing low prediction times.