Učenje modela za semantičku segmentaciju na nestandardnim skupovima podataka

master's thesis

• Title (english): Training semantic segmentation models on non-standard datasets
• Author: Ivan Šego
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Zoran Kalafatić (član povjerenstva)
• Committee member: Tomislav Hrkać (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-09-24, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Ovaj diplomski rad započinje uvodom o dubokom učenju i objašnjava zadatak seman-
tičke segmentacije na kojoj je rad baziran. U daljnjem poglavlju se opisuju osnovne
potrebne komponente dubokih konvolucijskih modela kako bi čitatelj imao potpuno
razumijevanje u nastavku teksta. Nadalje, daje se detaljan opis modela s gusto pove-
zanim putem za raspoznavanje i ljestvičastim naduzorkovanjem (ljestvičasto povezan
gusti model) koji se koristi za kao glavna arhitektura u radu. U idućem poglavlju da-
jemo pregled skupova koji se koriste za predtreniranje kodera i treniranje na zadatku
semantičke segmentacije. Nakon njega, dolazi najzanimljivije poglavlje gdje opisu-
jemo eksperimente i dajemo rezultate. Na kraju, dan je odgovarajući zaključak.

Abstract (english)

Master thesis starts with explaining deep learning and it also explains the purpose of semantic segmentation on which this thesis is based. In the next chapter, main component of deep convolutional model are explained so that reader can fully understand the rest of the thesis. After that, detailed description of Ladder-style DenseNet is given. Ladder-style DenseNet is the model which is used in this thesis. In the next chapter we have given description of the datasets which are used for experiments. After that, the most interesting part of this thesis is written and that is the report of the experiments and results. At the end, convenient conclusion is given.

• Keywords: Duboko učenje; Duboki konvolucijski modeli; Semantička segmentacija; Predtreniranje
• Keywords (english): Deep learning; Deep convolutional models; Semantic Segmentation; Pretraining
• Language: croatian
• URN:NBN: urn:nbn:hr:168:163040
• 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: 2023-01-13 13:54:43