Polunadzirana klasifikacija rukom pisanih znakova generativnim suparničkim modelima

master's thesis

• Title (english): Semi-supervised classification of hand-written characters by generative adversarial networks
• Author: Marko Jelavić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Marin Golub (član povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2017-07-20, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Klasifikacija slike je važno područje računalnog vida s velikim brojem zanimljivih primjena. Najbolji rezultati se postižu strogo nadziranim učenjem, no ručno označavanje podataka je izrazito skup proces. Kako bi smo smanjili potrebu za označavanjem služimo se polunadziranim učenjem, odnoso učenjem na djelomično označenim podacima. U ovom pristupu manji broj podataka je označen dok je većina neoznačena. Rad prezentira generativne suparničke modele koji se mogu koristiti u polunadzirane svrhe te razne implementacije istih i rezultate nad više skupova podataka.

Abstract (english)

Image classification is an important field of computer vision with many interesting applications. Best results are often accomplished with supervised learning, but manually labeling the data set is an expensive process. To lower the need for labeling big data sets, we use semi-supervised learning, which is basically learning on a partially labeled datas et. Smaller number of examples is labeled whereas bigger part of the data set is unlabeled. This thesis presents generative adversarial models which can be used in a semi-supervised setting along with many implementations of the aforementioned model. It also presents various implementations of the model and the experimental results on multiple data sets.

• Keywords: generativne suparničke mreže; polunadzirana klasifikacija
• Keywords (english): generative adversarial networks; semi-supervised classification
• Language: croatian
• URN:NBN: urn:nbn:hr:168:000802
• 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: 2019-03-13 21:11:24