Usporedba i evaluacija baza podataka emocija koristeći konvolucijsku neuronsku mrežu

master's thesis

• Title (english): Comparison and evaluation of emotion databases using a convolutional neural network
• Author: Marko Marijić
• Mentor: Tomislav Stipančić (mentor)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2022-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Mechanical Engineering

Abstract

Unutar ovog rada uspoređuju se dostupne baze podataka za automatsko prepoznavanje emocija s lica osobe. Zadatak je upotrebom VGG16 konvolucijske neuronske mreže provesti trening s tri različite baze podataka te evaluirati njihov rad. Koristeći isti model mreže, mogu se donijeti konkretni zaključci o kvaliteti podataka unutar svake baze. Uvodni dio rada pojašnjava što je to duboko učenje te predstavlja dosadašnja postignuća u području automatskog prepoznavanja emocija. Počevši od modela umjetnog neurona, objasniti će se cjelokupna arhitektura mreže. Definirati će se slojevi konvolucijske neuronske mreže, a zatim i objasniti svaki od njih. Nadalje, predstaviti će se baze podataka te opisati proces treninga. Na kraju će se prikazati dobiveni rezultati te će se komparativnom analizom donijeti zaključak o kvaliteti rada svake mreže. Za potrebe istraživanja u ovom radu korištene su brojne znanstvene metode od kojih se izdvajaju znanstvene metode analize i sinteze te metode indukcije i dedukcije.

Abstract (english)

Within this paper the available databases for automatic emotion recognition from facial features will be compared. The task is by using same VGG16 convolution neural network conduct training with three different databases and evaluate their work. Using the same network model, specific conclusions can be made about the quality of the date within each database. The introductory part of the paper explains what deep learning is and presents the achievements so far in the field of automatic emotion recognition. Starting with an artificial neuron model, the overall architecture of the network will be explained. The layers of convolutional neural networks will be defined and then each of them will be explained. Furthermore, databases will be presented, and the training process will be described. At the end, the obtained results will be presented, and a conclusion will be drawn about the quality of work of each network through a comparative analysis. For the purposes of research in this work, numerous scientific methods are used, of which the scientific methods of analysis and synthesis and the methods of induction and deduction stand out.

• Keywords: automatsko prepoznavanje emocija; prepoznavanje emocija s lica osobe; duboko učenje; VGG16; konvolucijska neuronska mreža
• Keywords (english): automatic emotion recognition; facial emotion recognition; FER; deep learning; VGG16; convolutional neural network; CNN
• Language: croatian
• URN:NBN: urn:nbn:hr:235:703511
• Study programme: Title: Mechanical Engineering; specializations in: Design, Process and Energy Engineering, Production Engineering, Engineering Modelling and Computer Simulation, Marine Engineering, Industrial Engineering and Management, Materials Engineering, Mechatronics and Robotics, Autonomous Systems; Course: Autonomous Systems; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka strojarstva (magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-11-30 20:29:58