Prepoznavanje starosti oseb s slik obrazov z uporabo konvolucijskih nevronskih mrež

Konda, Jaka (author), Peer, Peter (mentor)

Diplomsko delo pokaže celoten postopek razvoja rešitve problema prepoznavanja starosti oseb s slik. Začnemo s teoretičnimi osnovami o konvolucijskih nevronskih mrežah, s pomočjo katerih smo se tudi lotili problema. V praktičnem delu sledi priprava podatkov in učenje sestavljenega modela nevronske mreže, kjer smo izbrali znano VGG arhitekturo. Naučen model preizkusimo še na tekmovanju LAP, da dobimo rezultate, ki jih nato primerjamo z rešitvami ostalih ekip. Naši rezultati so se kljub nekoliko preprostejšemu pristopu izkazali za precej vzpodbudne, saj smo dosegli manjšo napako kot človek ter se uvrstili na 4. mesto med 11 ekipami.

• Language: Slovenian
• Keywords: računalniški vid, strojno učenje, nevronska mreža, konvolucijska nevronska mreža, klasifikacija, zaznavanje obrazov, starost, klasifikacija starosti, določanje starosti
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2016
• PID: 20.500.12556/RUL-84352
• Publication date in RUL: 16.08.2016

KONDA, Jaka, 2016, Prepoznavanje starosti oseb s slik obrazov z uporabo konvolucijskih nevronskih mrež [online]. Bachelor’s thesis. [Accessed 23 March 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=84352

Secondary language

• Language: English
• Title: Recognising people’s age from face images with convolutional neural networks

Abstract

The diploma thesis presents the entire process of developing a solution for recognising person’s age in the image. We start with the theoretical basics of convolutional neural networks that we used to address the problem. In the practical part we start with the preparation of used datasets and continue with learning of our neural network with the chosen widely known VGG architecture. Learned model is tested on the LAP competition dataset in order to obtain results, which are comparable with the solutions of other teams. Despite somewhat simpler approach our results proved to be quite encouraging. We surpassed human performance and ranked 4th among 11 teams.

• Keywords: computer vision, machine learning, neural networks, convolutional neural networks, classification, face detection, age, age classification, age recognition