Algoritmi za semantičku segmentaciju radioloških slika

undergraduate thesis

• Title (english): Algorithms for semantic segmentation of radilogical images
• Author: Petra Renić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Josip Šarić (član povjerenstva)
• Committee member: David Dukić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-07-06, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Semantička segmentacije predstavlja važan zadatak računalnog vida koji se primjenjuje u mnogim područjima. U posljednje vrijeme veliki napredak u tom zadatku postižu modeli koji se temelje na slojevima pažnje. Ovaj rad razmatra primjene semantičke segmentacije vezane uz analizu medicinskih slika. U okviru rada predstavljen je i opisan transformerski model SegFormer te je provedena evaluacija na Cityscapes skupu slika i učenje i evaluacija na Synapse skupu slika. Dobiveni rezultati su objašnjeni i uspoređeni sa rezultatima drugih radova koji se bave istim zadatkom. U radu su opisane osnove neuronskih mreža, kovolucije, transformerskih modela i slojeva pažnje. Opisan je i javno dostupan projekt koji je korišten za potrebe programske implementacije.

Abstract (english)

Semantic segmentation represents an important task in computer vision that is applied in various fields. Recently, significant progress has been achieved in this task using models based on attention layers. This paper discusses the applications of semantic segmentation related to the analysis of medical images. Within the paper, the SegFormer transformer model is presented and described, followed by an evaluation on the Cityscapes dataset and training and evaluation on the Synapse dataset. The obtained results are explained and compared with the results of other works that address the same task. The basics of neural networks, convolutions, transformer models, and attention layers are described in the paper. Additionally, a publicly available project used for the purpose of program implementation is mentioned.

• Keywords: računalni vid; semantička segmentacija; slojevi pažnje; SegFormer; transformer; kovolucijske neuronske mreže
• Keywords (english): computer vision; semantic segmentation; attention layers; SegFormer; transformer; convolutional neural networks
• Language: croatian
• URN:NBN: urn:nbn:hr:168:657181
• 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: 2023-10-11 10:19:51