Interaktivno okruženje za semantičku segmentaciju slika

undergraduate thesis

• Title (english): An Interactive framework for Semantic Image Segmentation
• Author: Josip Gatjal
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Mladen Karan (član povjerenstva)
• Committee member: Ivan Krešo (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2016-07-11, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Semantička segmentacija prirodnih scena je neriješen problem računalnog vida s mno-
gim zanimljivim primjenama. Posebno su zanimljivi strogo nadzirani pristupi kod kojih
je svaki odabir skupa za učenje označen pripadnim semantičkim razredom. Razmatran
je poseban oblik takvih pristupa u kojem se iterativno provodi interaktivno označavanje
slike i učenje modela, sve do postizanja željene subjektivne točnosti na razini cijele
slike. Za učenje modela zaslužne su duboke konvolucijske neuronske mreže koje se
tek nedavno počinju intenzivnije proučavati i ispitivati. Kao takve, podložne su mno-
gim promjenama i poboljšanjima što u konačnici može dovesti do boljih i konkretnijih
rezultata učenja mreže u budućnosti. Grafičko sučelje aplikacije omogućuje korisniku
laku i intuitivnu obradu željenih slika. Računalni vid iako već dosta istraženo područje
nastavlja evoluirati te se konstanto pronalaze nove ideje s kojima se želi iskoristiti mak-
simalni potencijal ove grane umjetne inteligencije. Grafičko sučelje je upravo jedna od
takvih ideja jer ima drugačiji pristup problemu semantičke segmentacije i time otvara
nove načine obrade i segmentiranja slika.

Abstract (english)

Semantic segmentation of natural scenes is an unsolved problem of computer vision with a lot of
interesting applications. Of particular interest are strictly controlled approaches in which
each selected set of learning is marked with the corresponding semantic class. Special form of such an approach is considered, in which we iteratively carried out interactive labeling
and learning of model image, until the desired accuracy of the subjective-wide
pictures is achieved. For learning the model we use deep convolution neural networks that are
only recently intensively studied and examined. As such, they are subject to much
changes and improvements which can ultimately lead to better and more concrete
results of learning networks in the future. The graphical interface allows the user application
easy and intuitive handling of the desired image. Computer vision, although already quite explored area,
continues to evolve and constantly find new ideas that take advantage of maximum
potential of this branch of artificial intelligence. The graphical interface is just one of
such ideas because it has a different approach to the problem of semantic segmentation and thus opens
new ways of processing and segmentation of images.

• Keywords: semantička segmentacija; neruonske mreže; grafičko sučelje; python; tkinter
• Keywords (english): semantic segmentation; neural network; graphical interface; python; tkinter
• Language: croatian
• URN:NBN: urn:nbn:hr:168:878293
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2018-02-06 08:51:36