Raspoznavanje izraza lica temeljeno na dubokom učenju

master's thesis

• Title (english): Facial Expression Recognition Based on Deep Learning
• Author: Franko Šesto
• Mentor: Zoran Kalafatić (mentor)
• Committee member: Zoran Kalafatić (predsjednik povjerenstva)
• Committee member: Siniša Šegvić (član povjerenstva)
• Committee member: Tomislav Hrkać (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-11, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Analiza i raspoznavanje izraza lica je jedno od trenutno popularnih područja koje ima zanimljive primjene u sustavima za interakciju čovjeka i računala. U ovom radu fokus je na dva bitna dijela koji određuju kvalitetu sustava: praćenju lica te klasifikaciji emocija izraženih ekspresijama lica. Praćenje lica je izazovan problem zbog velike različitosti lica i ekspresija lica te različitosti okolina u kojima se ona nalaze. Rješenje je teško dobiti zbog varijacija u osvijetljenosti, perspektivi lica te ekspresiji lica u različitim okolinama. Opisani su raznovrsni algoritmi u tom području te su iskorišteni za dobivanje dobrog praćenja kao potrebnog dijela sustava. S druge strane, za raspoznavanje emocija se koriste konvolucijske neuronske mreže. Opisano je kako funkcioniraju i koja su njihova dobra svojstva kao i dobri obrasci za dizajn arhitektura i treniranje mreža. Razmotrene su korištene arhitekture kao i dobivene točnosti na korištenom skupu slika. Na kraju su oba dijela ukomponirana u sustav koji raspoznaje emocije čovjeka u stvarnom vremenu.

Abstract (english)

Facial expression recognition and analysis is one of the currently popular areas that has interesting applications in human-computer interaction. This thesis is focused on two essential parts which decide system quality: face tracking and emotion classification based on facial expressions. Face tracking is a challenging problem, due to the large variability of faces and facial expressions, and context where they can be found. It is very difficult to come up with good solution due to the light variations, pose, different face expressions etc. Various algorithms are described from this area and they are used to get good tracking as necessary part of this system. On the other hand, convolutional neural networks are used for emotion recognition. It is described how they function and what are their good properties and good patterns for architecture design and network training. Various network architectures are trained on a large dataset and accuracies for those architectures are obtained. Both parts are combined into system able to recognize human emotions in real time.

• Keywords: detekcija lica; praćenje lica; klasifikacija emocija; duboko učenje; konvolucijske neuronske mreže; algoritam Viole i Jonesa; podudaranje predloškom; optički tok
• Keywords (english): face detection; face tracking; emotion classification; deep learning; convolutional neural networks; Viola-Jones algorithm; template matching; optical flow
• Language: croatian
• URN:NBN: urn:nbn:hr:168:228203
• 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: 2020-01-07 22:20:01