Segmentacija sastojaka na slikama hrane

undergraduate thesis

• Title (english): Ingredient segmentation in food images
• Author: Bruno Mikulan
• Mentor: Sonja Grgić (mentor)
• Committee member: Sonja Grgić (predsjednik povjerenstva)
• Committee member: Jelena Božek (član povjerenstva)
• Committee member: Ivana Žeger (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-05, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Ovaj rad istražuje primjenu metoda segmentacije hrane na slikama kako bi se omogućila precizna identifikacija i izolacija pojedinih komponenti hrane. Korištene su metode K-srednjih vrijednosti i razvodnice uz upotrebu ResNet neuronskih mreža za klasifikaciju. Rezultati pokazuju da metoda razvodnice pruža bolje rezultate u odnosu na K-srednje vrijednosti, iako obje metode imaju izazove s presegmentacijom ili podsegmentacijom. Klasifikacija pomoću ResNet18 modela postigla je točnost od 79% bez prethodne segmentacije. Prepoznaje se potreba za daljnjim istraživanjem kako bi se razvile optimizirane metode segmentacije hrane i unaprijedile tehnike klasifikacije.

Abstract (english)

This paper explores the application of food segmentation methods in images to facilitate precise identification and isolation of individual food components. Both K-means clustering, and watershed methods were employed alongside ResNet neural networks for classification. Results indicate that the watershed method outperforms K-means clustering, although both methods face challenges with over-segmentation or under-segmentation. Classification using the ResNet18 model achieved an accuracy of 79% without prior segmentation. There is a recognized need for further research to develop optimized food segmentation methods and advance classification techniques.

• Keywords: segmentacija; računalni vid; duboko učenje; ResNet; prepoznavanje hrane; K-srednjih vrijednosti; metoda razvodnice; Food-101
• Keywords (english): K-means clustering; Watershed method; Neural networks; ResNet; Deep learning; Food recognition; Computer vision; Clustering; Image masking; Food-101 dataset
• Language: croatian
• URN:NBN: urn:nbn:hr:168:663864
• 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: 2024-10-12 21:35:24