Primjena alata dubokog učenja za prepoznavanje lica

undergraduate thesis

• Title (english): Application of deep learning tools for face recognition
• Author: Karmen Santini
• Mentor: Sanja Seljan (mentor)
• Committee member: Ivan Dunđer (predsjednik povjerenstva)
• Committee member: Vjera Lopina (član povjerenstva)
• Committee member: Sanja Seljan (član povjerenstva)
• Granter: University of Zagreb; Faculty of Humanities and Social Sciences; (Department of information and Communication sciences); Zagreb
• Defense date and country: 2021-09-20, Croatia
• Scientific / art field, discipline and subdiscipline: SOCIAL SCIENCES; Information and Communication Sciences

Abstract

Cilj ovog rada bio je prikazati postupak treniranja, prepoznavanja i zamjene lica primjenom alata za duboko učenje, koje predstavljaju najsuvremeniji pristup u području umjetne inteligencije i primjena neuronskih mreža prilikom učenja i definiranja lica, te ukazati na potencijalne stvarne prijetnje korištenja takve tehnologije, ali i pozitivne strane.
Rad se sastoji od teorijskog i praktičnog dijela. U teorijskom dijelu definirani su pojmovi strojno učenje i duboko učenje i ključni elementi i razlike. Prikazana je struktura i način rada neuronske mreže, te osnovni principi rada koji se koriste za duboko učenje. Naglasak rada je na primjenama dubokog učenja te su prikazane tehnike koje koriste razne aplikacije u kojima je implementirano duboko učenje za prepoznavanje lica.
U praktičnom dijelu rada provedeno je istraživanje primjenom alata za prepoznavanje lica. Računalni program FaceSwap primjenjuje princip dubokog učenja za stvaranje „dubokih lažnjaka“ (eng. deepfakes) u kojima algoritam uči glavne značajke lica te izvorna lica na slikama ili videozapisima zamjenjuje željenim licem neke druge osobe. Nakon prikaza sustava i analize načina rada, prikazan je praktičan dio zamjene lica u videozapisu te su na kraju prikazani rezultati. Posljednje poglavlje odnosi se na etičke principe i ukazuje na opasnosti i moguću zlouporabu te zaštitu od neprimjerene upotrebe.

Abstract (english)

The aim of this paper was to present the process of training, recognizing and replacing faces using deep learning tools, which represent the most modern approach in the field of artificial intelligence and neural networks in learning and defining faces, and point out potential real threats of using such technology. sides.
The paper consists of a theoretical and a practical part. In the theoretical part, the terms machine learning and deep learning and key elements and differences are defined. The structure and mode of operation of the neural network are presented, as well as the basic principles of operation used for deep learning. The emphasis of the paper is on the applications of deep learning and the techniques used by various applications in which deep learning for facial recognition is implemented are presented.
In the practical part of the paper, research was conducted using facial recognition tools. The computer program FaceSwap applies the principle of deep learning to create "deep fakes" in which the algorithm learns the main features of the face and replaces the original faces in pictures or videos with the desired face of another person. After showing the system and analyzing the mode, the practical part of the face replacement in the video is shown and finally the results are shown. The last chapter deals with ethical principles and points out the dangers and possible abuse and protection against inappropriate use.

• Keywords: duboko učenje; primjena; neuronske mreže; strojno učenje; učenje; prepoznavanje lica
• Keywords (english): deep learning; application; neural networks; machine learning; learning; facial recognition
• Language: croatian
• URN:NBN: urn:nbn:hr:131:991319
• Study programme: Title: Information Sciences; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/prvostupnica (baccalaureus/baccalaurea) informacijskih znanosti (sveučilišni/a prvostupnik/prvostupnica (baccalaureus/baccalaurea) informacijskih znanosti)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2021-09-20 15:56:56