Renderiranje scene pomoću dubokog učenja

master's thesis

• Title (english): Deep learning approach to scene rendering
• Author: Niko Pave Žagar
• Mentor: Marija Seder (mentor)
• Committee member: Ivan Marković (predsjednik povjerenstva)
• Committee member: Luka Petrović (član povjerenstva)
• Committee member: Juraj Peršić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-09-21, Croatia

Abstract

Kroz ovaj rad prikazan je proces rekonstrukcije trodimenzionalnog modela iz 2D slike te njegovo renderiranje. U početku su objašnjeni osnovni koncepti oblaka točaka, estimacije dubine, registracije i 3D rekonstrukcije. Prvi korak je estimacija dubine iz dvodimenzionalne slike i stvaranja udbinske mape, iz koje se generira oblak točaka. Ispitane su tri različite metode registracije oblaka točaka, od kojih su samo ICP algoritmom dobiveni rezultati, dok za TEASER i PREDATOR metode oblaci točaka generirani estimacijom nisu valjani. Dobiveni oblaci točaka rekonstruirani su Poissonovom metodom rekonstrukcije površine. Objašnjeni su temeljni principi renderiranja te su u konačnici prikazani rezultati dobivenih testiranja.

Abstract (english)

Through this work, the process of reconstructing a three-dimensional model from a 2D image and its rendering is presented. Initially, the fundamental concepts of point clouds, depth estimation, registration, and 3D reconstruction are explained. The first step involves depth estimation from a two-dimensional image and the creation of a depth map, from which a point cloud is generated. Three different methods for point cloud registration were tested, with valid results obtained only using the ICP algorithm, while point clouds couldn't be generated by TEASER and PREDATOR methods. The obtained point clouds were reconstructed using the Poisson surface reconstruction method. The basic principles of rendering are explained, and the results of the conducted tests are ultimately presented.

• Keywords: Oblak točaka; registracija oblaka točaka; 3D rekonstrukcija; ICP; TEASER; PREDATOR; Poissonova metoda rekonstrukcije površine; renderiranje
• Keywords (english): Point cloud; point cloud registration; 3D reconstruction; ICP; TEASER; PREDATOR; Poisson surface reconstruction method; rendering
• Language: croatian
• URN:NBN: urn:nbn:hr:168:339638
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-10-11 11:04:46