magistrsko delo
Gregor Horvat (Author), Štefan Kohek (Mentor), Aljaž Jeromel (Co-mentor)

Abstract

V magistrskem delu sta opisani zasnovi dveh pristopov strojnega učenja za razpoznavo značilnosti v človeškem očesu na podlagi majhne učne množice. Implementirani sta dve tehniki; semantična segmentacija in lokalizacija. Obe rešitvi delujeta na osnovi konvolucijskih nevronskih mrež in iz digitalnih fotografij očes razpoznata položaj zenice, zunanjo obrobo šarenice ter barvo le-te. Problem omejene učne množice smo naslovili z uporabo več tehnik obogatitve učne množice na podlagi obstoječih učnih podatkov. Najboljše rezultate je dosegla segmentacijska nevronska mreža, tehnike obogatitve učne množice pa so se izkazale za nepogrešljive pri učenju na majhni učni množici.

Keywords

konvolucijske nevronske mreže;semantična segmentacija;lokalizacija;obogatitev učne množice;magistrske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.09 - Master's Thesis
Organization: UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher: [G. Horvat]
UDC: 004.93'1:004.85(043.2)
COBISS: 149194499 Link will open in a new window
Views: 108
Downloads: 15
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Automatic detection of eye features using machine learning with a small dataset
Secondary abstract: In this master's thesis we designed two different machine learning approaches for detection of eye features based on a small training dataset. Implemented and described are two solutions; semantic segmentation and localization. Both are based on convolutional neural networks, and are able to detect iris position, iris contour and it's color from digital images. We addressed the problem of a limited training dataset with multiple augmentation techniques, which work with existing data. The best results were achieved with semantic segmentation approach. Data augmentation techniques proved to be an essential tool when working with a limited data set.
Secondary keywords: convolutional neural networks;semantic segmentation;localization;dataset expansion;
Type (COBISS): Master's thesis/paper
Thesis comment: Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Telekomunikacije
Pages: 1 spletni vir (1 datoteka PDF (XI, 49 f.))
ID: 17902916