prikaz prve stranice dokumenta Samo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju
No public access
master's thesis
Samo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju
Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing, 2022. urn:nbn:hr:168:092234
Kapusović, Karlo
University of Zagreb
Faculty of Electrical Engineering and Computing

Institutional repository: FER Repository

Cite this document

Kapusović, K. (2022). Samo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju (Master's thesis). Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing. Retrieved from https://urn.nsk.hr/urn:nbn:hr:168:092234

Kapusović, Karlo. "Samo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju." Master's thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2022. https://urn.nsk.hr/urn:nbn:hr:168:092234

Kapusović, Karlo. "Samo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju." Master's thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2022. https://urn.nsk.hr/urn:nbn:hr:168:092234

Kapusović, K. (2022). 'Samo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju', Master's thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, accessed 12 July 2024, https://urn.nsk.hr/urn:nbn:hr:168:092234

Kapusović K. Samo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju [Master's thesis]. Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing; 2022 [cited 2024 July 12] Available at: https://urn.nsk.hr/urn:nbn:hr:168:092234

K. Kapusović, "Samo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju", Master's thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, 2022. Available at: https://urn.nsk.hr/urn:nbn:hr:168:092234

Metadata
TitleSamo-nadzirano učenje dubine za monokularnu direktnu vizualnu odometriju
Title (english)Self-Supervised Depth Learning for Monocular Direct Visual Odometry
AuthorKarlo Kapusović
MentorIvan Marković (mentor)
Committee memberIvan Petrović (član povjerenstva)
Committee memberMarija Seder (član povjerenstva)
Committee memberJuraj Peršić (član povjerenstva)
GranterUniversity of Zagreb
Faculty of Electrical Engineering and Computing
Zagreb
Defense date and country2022-07-13, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Electrical Engineering
Abstract
Direktna vizualna odometrija skupina je algoritama koji estimiraju položaj kamere pomoću vrijednosti intenziteta piksela u slikama. Samim time je osjetljiva na pogreške kalibracije i inicijalizacije početnih vrijednosti. Iako stereo kamere omogućuju brzu inicijalizaciju i procjenu metričke skale, njihovo korištenje je složenije i zahtijeva kompleksniju kalibraciju kamera. Duboko učenje omogućuje estimaciju metrički točne dubine iz monokularnih slika. Korištenjem najnovijih pristupa samo-nadziranog učenja, omogućuje se učenje mreže bez označenog skupa podataka za treniranje. Stoga je moguća vrlo laka prilagodba neuronske mreže za različita okruženja. U diplomskome radu potrebno je iskoristiti dubinske mape iz neuronske mreže u direktnoj vizualnoj odometriji kako bi se poboljšala inicijalizacije i omogućila procjena metričke skale svijeta iz monokularnih slika.
Abstract (english)
Direct visual odometry is a group of algorithms that estimate camera position using pixel intensity values in images. It is therefore sensitive to calibration errors and initialization of initial values. Although stereo cameras allow for rapid initialization and metric scale estimation, their use is more complex and requires more complex camera calibration. Deep learning allows the estimation of metrically accurate depth from monocular images. Using the latest approaches to self-supervised learning, network learning is enabled without a labeled training data set. Therefore, it is possible to very easily adapt the neural network to work in different environments. In this graduate thesis it is necessary to use depth maps from the neural network in direct visual odometry in order to improve the initializations and enable the assessment of the metric scale of the world from monocular images.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:168:092234
Study programmeTitle: Electrical Engineering and Information Technology
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra inženjer/inženjerka elektrotehnike i informacijske tehnologije (magistar/magistra inženjer/inženjerka elektrotehnike i informacijske tehnologije)
Type of resourceText
File originBorn digital
Access conditionsClosed access
Terms of useLicence In copyright icon
Public note
Created on2023-01-13 15:15:19