Duboko učenje za vizualnu odometriju

undergraduate thesis

• Title (english): Deep learning for visual odometry
• Author: Jelena Lončar
• Mentor: Ivan Marković (mentor)
• Committee member: Ivan Marković (član povjerenstva)
• Committee member: Marta Čolaković-Bencerić (član povjerenstva)
• Committee member: Vlaho-Josip Štironja (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

Rad istražuje implementaciju i evaluaciju sustava vizualne odometrije temeljenog na dubokom učenju, poznatog kao DeepVO. Vizualna odometrija ključna je značajka za praćenje kretanja kamere u stvarnom vremenu. Naglasak je na integraciji dubokih neuronskih mreža u svrhu poboljšanja preciznosti i robusnosti odometrijskog sustava automatskim ekstrahiranjen složenih značajki iz vizualnih podataka. Pomoću detaljno opisanih metodoloških pristupa, predočava se arhitektura DeepVO sustava, implementirana putem PyTorch okvira za duboko učenje. Evaluacija performansi izvršena je na relevantnim skupovima podataka koja zaključuje rad diskusijom o postignutim rezultatima i mogućim smjerovima budućih istraživanja.

Abstract (english)

The paper analyzes the implementation and evaluation of the visual odometry system based on deep learning, also known as DeepVO. Visual odometry is the key component for tracking videocamera movement in real time. The emphasis is on the integration of deep neural networks which improve the precision and robustness of odometric systems by automatically extracting complex features from visual data. Using detailed methodological approaches, the architecture of the DeepVO system, implemented through the PyTorch framework for deep learning, is presented. The benchmark testing is performed on the relevant data sets. It concludes the paper with a discussion of the results achieved and possible directions for future research.

• Keywords: Vizualna odometrija; duboko učenje; DeepVO; konvolucijske neuronske mreže; rekurentne neuronske mreže; KITTI Vision Benchmark Suite
• Keywords (english): Visual odometry; deep learning; DeepVO; convolutional neural networks; recurrent neural networks; KITTI Vision Benchmark Suite
• Language: croatian
• URN:NBN: urn:nbn:hr:168:054946
• 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 21:20:24