Semantička segmentacija harmoničkim gusto povezanim modelima

undergraduate thesis

• Title (english): Semantic segmentation with harmonic densely connected models
• Author: Anja Delić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Petra Bevandić (član povjerenstva)
• Committee member: Josip Šarić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-09, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Jedan od mnogih zanimljivih zadataka računalnog vida je semantička segmentacija. Modeli s gusto povezanim konvolucijskim slojevima postižu vrlo dobre rezultate, a u posljednje vrijeme pojavila su se poboljšanja koja pažljivim povezivanjem konvolucijskih slojeva nastoje smanjiti broj memorijskih zahtjevnih operacija tijekom učenja i testiranja bez smanjenja točnosti modela. U ovom radu predstavljene su harmoničke gusto povezane konvolucijske mreže te je provedeno učenje i evaluacija modela FCHarDNet70, jedne konkretne implementacije harmoničkog gusto povezanog modela. Rezultati su prikazani i objašnjeni na kraju rada. Dodatno, ukratko su opisane osnove dubokog učenja i konvolucijskih mreža. Opisan je i postupak učenja, korišteni skupovi podataka, detalji programske implementacije i korištene biblioteke.

Abstract (english)

One of the many interesting tasks of computer vision is semantic segmentation. Models with densely connected convolutional layers achieve very good results, and recently there have been improvements that, by carefully connecting convolutional layers, seek to reduce the number of memory-demanding operations during learning and testing without reducing model accuracy. In this paper, harmonic densely connected convolutional networks are presented and the learning and evaluation of the FCHarDNet70 model, a specific implementation of a harmonic densely connected model, is performed. The results are presented and explained at the end of the paper. Additionally, the basics of deep learning and convolutional networks are briefly described. The learning process, used data sets, details of program implementation and used libraries are also described.

• Keywords: računalni vid; semantička segmentacija; konvolucijske neuronske mreže; harmonički gusto povezani modeli
• Keywords (english): computer vision; semantic segmentation; convolutional neural networks; harmonic densley connected models
• Language: croatian
• URN:NBN: urn:nbn:hr:168:033039
• 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: 2023-01-14 20:17:27