Lokalizacija objekata u slikama korištenjem dubokog učenja

undergraduate thesis

• Title (english): Object localization in images using deep learning
• Author: Dominik Barukčić
• Mentor: Tomislav Hrkać (mentor)
• Committee member: Tomislav Hrkać (predsjednik povjerenstva)
• Committee member: Zorana Doždor (član povjerenstva)
• Committee member: Iva Sović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-02, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Rad istražuje problem lokalizacije i klasifikacije rukom pisanih znamenki primjenom metoda dubokog učenja. Cilj je prilagoditi skup podataka, razviti precizan model za klasifikaciju i lokalizaciju, naučiti ga i evaluirati. Skup podataka MNIST prilagođen je zahtjevima modela, a prije treniranja potrebno je obaviti pretprocesiranje ulaznog skupa podataka. Model je temeljen na konvolucijskim neuronskim mrežama, koje su učinkovite za ovakve zadatke. Evaluacija je provedena upotrebom metrika poput... More matrice konfuzije i Jaccardovog indeksa, što omogućuje detaljnu analizu performansi modela. Model je postigao točnost klasifikacije od 98.35% i točnost lokalizacije od 94.14%. U radu su predstavljeni rezultati detekcije znamenki te rasprava o kvaliteti i mogućim izazovima, kao i prijedlozi za buduća poboljšanja. Less

Abstract (english)

The paper investigates the problem of localization and classification of handwritten digits using deep learning methods. The objective is to adapt the dataset, develop an accurate model for classification and localization, train it, and evaluate its performance. The MNIST dataset was adapted to meet the model's requirements, and preprocessing of the input dataset was necessary before training. The model is based on convolutional neural networks, which are effective for such tasks. Evaluation... More was conducted using metrics such as the confusion matrix and the Jaccard index, enabling a detailed analysis of the model's performance. The model achieved a classification accuracy of 98.35% and a localization accuracy of 94.14%. The paper presents the results of digit detection, discusses the quality and potential challenges, and provides suggestions for future improvements. Less

• Keywords: duboko učenje; klasifikacija; lokalizacija; MNIST; konvolucijske neuronske mreže; funkcije gubitka; računalni vid
• Keywords (english): deep learning; classification; localization; MNIST; convolutional neural networks; loss functions; computer vision
• Language: croatian
• URN:NBN: urn:nbn:hr:168:953599
• 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: Open access
• Created on: 2024-10-12 20:04:13