Raspoznavanje objekata dubokim neuronskim mrežama

master's thesis

• Title (english): Object recognition by deep neural networks
• Author: Vedran Vukotić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Jan Šnajder (član povjerenstva)
• Committee member: Marko Čupić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2014-07-17, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom su radu prikazani načini rada i implementacijski detalji dubokih neuronskih mreža sa fokusom na konvolucijske neuronske mreže. Napravljen je sustav za brzu izradu različitih konvolucijskih neuronskih mreža i njihovo testiranje. Testirane su dvije različite konvolucijske neuronske mreže: jedna za klasifikaciju podataka iz skupa rukom pisanih znamenki MNIST te druga za prepoznavanje hrvatskih prometnih znakova iz skupa FER-MASTIF TS2010. Na kraju su opisani dobiveni rezultati te su analizirani krivo prepoznati uzorci.

Abstract (english)

In this work, the specific mechanisms and implementational details of deep neural networks are shown with an emphasis on convolutional neural networks. A system for fast convolutional neural network prototyping is made and two different CNNs are tested: one for handwritten digits recognition, based on the MNIST dataset, and the second for Croatian traffic signs recognition, based on the FER-MASTIF TS2010 dataset. Following that, numerical results are expressed from the classifiers, their performance is evaluated and misclassified examples are briefly discussed.

• Keywords: neuronske mreže; duboke neuronske mreže; konvolucijske neuronske mreže; višeklasna klasifikacija; raspoznavanje prometnih znakova; FER-MASTIF TS2010; MNIST; računalni vid; raspoznavanje uzoraka; duboko učenje
• Keywords (english): neural networks; deep neural networks; convolutional neural networks; multiclass classification; traffic signs recognition; FER-MASTIF TS2010; MNIST; computer vision; pattern recognition; deep learning
• Language: croatian
• URN:NBN: urn:nbn:hr:168:749831
• 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: 2018-02-06 14:29:02