Semantička segmentacija zgrada u satelitskim snimkama

undergraduate thesis

• Title (english): Semantic segmentation of buildings in satellite imagery
• Author: Gabrijel Jambrošić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Marin Kačan (član povjerenstva)
• Committee member: Matej Grcić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-06, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Ovaj rad bavi se problemom semantičke segmentacije zgrada u satelitskim snimkama. Rezidualne mreže općenito su se pokazale kao dobro rješenje semantičke segmentacije. Zbog toga je za ovaj rad korištena arhitektura SwiftNet s piramidalnom fuzijom, inače stvorena za semantičku segmentaciju u stvarnom vremenu. Ta arhitektura pokazala je vrlo dobre rezultate za segmentaciju snimki u prometu, a jedna od prednosti joj je brzo učenje. Slike satelitskih snimki dolaze iz skupa podataka Inria, od čega je dio iskorišten za treniranje, a dio za validaciju modela. Sve slike smanjene su na pola rezolucije kako bih zadovoljio resursna ograničenja. U radu su objašnjeni osnovni pojmovi potrebni za shvaćanje korištenih metoda nakon čega su izneseni dobiveni rezultati.

Abstract (english)

This paper deals with the problem of semantic segmentation of buildings in satellite imagery. Residual networks have generally proven to be a good solution for semantic segmentation. Therefore, the SwiftNet architecture with pyramidal fusion, otherwise created for real-time semantic segmentation, was used for this paper. This architecture has shown very good results for segmentation of traffic images, and one of its advantages is fast learning. Satellite images come from the Inria data set, part of which was used for training, and part for validating the model. All images are reduced to half resolution in order to meet computing resource constraints. The paper explains the basic terms needed to understand the methods used after which the obtained results are presented.

• Keywords: semantička segmentacija; SwiftNet; Inria; piramidalna fuzija
• Keywords (english): semantic segmentation; SwiftNet; Inria; pyramidal fusion
• Language: croatian
• URN:NBN: urn:nbn:hr:168:925294
• 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: 2023-01-15 10:23:39