Detekcija zuba u ortopantomogramima

undergraduate thesis

• Title (english): Detection of Teeth in Orthopantomograms
• Author: Miljenko Šuflaj
• Mentor: Marko Subašić (mentor)
• Committee member: Marko Subašić (predsjednik povjerenstva)
• Committee member: Ivana Zadro Matovinović (član povjerenstva)
• Committee member: Karlo Koščević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Metodama dubokog učenja uz arhitekturu RetinaNet istražujemo jednoprolaznu neuronsku mrežu i njen učinak na problemu detekcije zuba u ortopantomogramima. Prvo se upoznajemo s najosnovnijim konceptima neuronskih mreža općenito, pa krećemo u specifičnosti našeg zadatka. Uspoređujemo različite pristupe problemu uz bazu podataka za učenje koja sadrži zube koji pripadaju širokom rasponu ljudi. Dolazimo do optimalne srednje preciznosti od 90.26%, što je nakon usporedbe bolje od ijednog rezultata srodnih radova do sad. Tijekom svega ovoga razmatramo o stvarima koje su nas sputavale ili o pretpostavkama kada ne znamo sigurno. Pronalazimo potencijalna rješenja za neke od problema s kojima se susrećemo. Na kraju sagledavamo rješavanje ovog problema s pogleda performansi.

Abstract (english)

Using deep learning methods and the RetinaNet architecture we explore a single pass neural network and it's efficiency regarding the teeth detection problem. The first step is understanding the most basic neural network concepts after which we delve into the specifics of our task. We compare different approaches to our problem using our training database which contains teeth belonging to a wide spectrum of people. We derive the optimal average precision of 90.26%, which ends out being the best one when compared to all related scientific work in the same faculty. During all of this we discuss things that may have slowed us down or think about it when we're not exactly sure where the problem lies. We find potential solutions for some of the problems we face. At the end we look at this problem from a performance perspective.

• Keywords: duboko učenje; neuronske mreže; ortopantomogrami; RetinaNet; žarišni gubitak
• Keywords (english): deep learning; neural networks; orthopantomograms; RetinaNet; focal loss
• Language: croatian
• URN:NBN: urn:nbn:hr:168:763280
• 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: 2020-01-07 22:32:59