Semantička segmentacija slika dubokim konvolucijskim mrežama na mobilnim uređajima

master's thesis

• Title (english): Semantic Segmentation of Images by Deep Convolutional Networks on Mobile Devices
• Author: Ante Prpić
• Mentor: Vladimir Čeperić (mentor)
• Committee member: Vladimir Čeperić (predsjednik povjerenstva)
• Committee member: Siniša Šegvić (član povjerenstva)
• Committee member: Zoran Kalafatić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-15, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Diplomskim radom opisan je problem semantičke segmentacije prirodnih scena riješen dubokim konvolucijskim mrežama koje su detaljno opisane. Objašnjeno je kako neuronska mreža uči te kako konvolucija utječe na rješavanje problema iz domene računalnog vida. Evaluirane su popularne arhitekture semantičke segmentacije U-Net i DeepLab te je naučen model na CamVid skupu podataka koristeći DeepLab arhitekturu. Demonstrira se prevođenje modela u strukturu podataka pogodnu za mobilnu aplikaciju te se implementira mobilna iOS aplikacija s mogućnošću segmentacije scene u realnom vremenu čija funkcionalnost je analizirana na više mobilnih uređaja različitih generacija.

Abstract (english)

The diploma thesis describes the problem of semantic segmentation of natural scenes solved by deep convolutional networks which are described in detail. It is explained how the neural network learns and how convolution helps solve problems from the domain of computer vision. Popular U-Net and DeepLab semantic segmentation architectures were evaluated and a model was learned on a CamVid dataset using the DeepLab architecture. The translation of the model into a data structure suitable for a mobile application is demonstrated, and a mobile iOS application with the possibility of real-time scene segmentation is implemented, the functionality of which has been analyzed on several mobile devices of different generations.

• Keywords: konvolucijska neuronska mreža; semantička segmentacija; konvolucija; dilatirana konvolucija; duboko učenje; CoreML; računalni vid; DeepLab; MobileNet
• Keywords (english): convolutional neural network; semantic segmentation; convolution; dilated convolution; deep learning; CoreML; computer vision; DeepLab; MobileNet
• Language: croatian
• URN:NBN: urn:nbn:hr:168:195659
• 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 13:35:18