Autentikacija osoba analizom izgleda lica

master's thesis

• Title (english): Person authentication by face appearance analysis
• Author: Ivana Babić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Zoran Kalafatić (član povjerenstva)
• Committee member: Vladimir Čeperić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-09-24, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U sklopu ovog rada proučene su i opisane arhitekture konvolucijskih neuronskih mreža i funkcije gubitka za problem raspoznavanje lica i prepoznavanja izraza lica. Izgrađen je sustav za autentikaciju lica koji se sastoji od komponenti za detekciju lica, detekciju treptaja, modela za raspoznavanje ekspresija lica učenog na FER2013 skupu podataka te evaluiranom na CK+48 skupu podataka, klasifikatora ExtraTreesClassifier učenog na LCC-FASD skupu podataka i modela za raspoznavanje lica na otvorenom skupu učenog na MS1M-ArcFace skupu podataka i evaluiranom na LFW skupu podataka.
Prikazani su i ocijenjeni rezultati svih implementiranih modela, prikazani pogrešno klasificirani primjeri te je provedena detaljna usporedba s rezultatima iz literature.
Sve komponente povezane su u sustav za autentikaciju te je, zajedno s korisničkim sučeljem implementiranim pomoću Kivy biblioteke, predstavljena funkcionalna aplikacija za autentikaciju osoba u stvarnom vremenu.

Abstract (english)

In this work, the specific architectures and loss functions for deep neural networks are shown with an emphasis on convolutional neural networks for face recognition and facial expression recognition. A system for face authentication has been implemented, containing components for face detection, blink detection, model for face expression detection trained on FER2013 dataset and evaluated on CK+48 dataset, ExtraTreesClassifier trained on LCC-FASD dataset and model for face recognition trained on MS1M-ArcFace dataset and evaluated on LFW dataset.

Furthermore, numerical results of all implemented models are expressed, their performance is evaluated and compared with the results from the literature and misclassified examples are briefly discussed.
All mentioned components are connected together to form a functional real-time face authentication system with user friendly interface implemented using Kivy framework.

• Keywords: neuronske mreže; duboke neuronske mreže; konvolucijske neuronske mreže; raspoznavanje lica; računalni vid; duboko učenje; ResNet; ResNet50; Xception; Mini_Xception; ExtraTreesClassifier; ArcFace; SphereFace; MSM1-Arcface; LFW; FER2013; CK+48; LCC-FASD; raspoznavanje izraza lica; detekcija treptanja; prezentacijski napadi
• Keywords (english): neural networks; deep neural networks; convolutional neural networks; face recognition; computer vision; deep learning; ResNet; ResNet50; Xception; Mini_Xception; ExtraTreesClassifier; ArcFace; SphereFace; MSM1-Arcface; LFW; FER2013; CK+48; LCC-FASD; facial expression recognition; blink detection; presentation attacks
• Language: croatian
• URN:NBN: urn:nbn:hr:168:014013
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 09:03:22