Rekonstrukcija elipsoidov iz ene same slike z uporabo konvolucijske nevronske mreže : magistrsko delo

Ivanšek, Rok (author), Solina, Franc (mentor)

V zadnjih letih je zaradi tehnoloških napredkov, hitro dostopne in poceni procesorske moči in velike količine podatkov, tehnologija globokih nevronskih mrež doživela preporod. Med drugim se uporablja na področju avtonomne vožnje, strojnih prevodov, prepoznavanja slik in zdravstvu. V tej magistrski nalogi uporabimo posebno vrsto globokih nevronskih mrež, z imenom konvolucijske nevronske mreže, na konkretnem problemu iz področja računalniškega vida. Natančneje, z implementacijo tro-faznega modela, ki temelji na tehnologiji konvolucijskih nevronskih mrež, predstavimo avtomatizirano in hitro metodo za rekonstrukcijo elipsoidov iz ene same barvne slike.

• Language: Slovenian
• Keywords: globoke nevronske mreže, konvolucijske nevronske mreže, računalniški vid, rekonstrukcija elipsoidov, rekonstrukcija volumetričnih modelov iz ene same slike
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FMF - Faculty of Mathematics and Physics; FRI - Faculty of Computer and Information Science
• Year: 2019
• PID: 20.500.12556/RUL-108571
• UDC: 004
• COBISS.SI-ID: 18669913
• Publication date in RUL: 07.07.2019

IVANŠEK, Rok, 2019, Rekonstrukcija elipsoidov iz ene same slike z uporabo konvolucijske nevronske mreže : magistrsko delo [online]. Master’s thesis. [Accessed 23 March 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=108571

Secondary language

• Language: English
• Title: Using convolutional neural networks to reconstruct ellipsoids out of a single image

Abstract

In the last few years one can observe a rise and rebirth of deep neural networks. This is mostly due to the availability of big data and technological advances in computer hardware that made processing power cheap and easily accessible. Deep neural networks are being used in several application domains, for example in autonomous vehicles, for machine translation, image recognition and in health care. In this master thesis we use convolutional neural networks, a flavour of deep neural networks, to solve a problem in the domain of computer vision. By implementing a three-phase model that is based on convolutional neural networks we propose a fast and automated method for reconstructing ellipsoids out of a single color image.

• Keywords: deep neural networks, convolutional neural networks, computer vision, ellipsoid reconstruction, 3D model reconstruction from a single image