Estimacija dispariteta iz stereo slika zasnovana na dubokim neuronskim mrežama

master's thesis

• Title (english): Disparity estimation from stereo images based on deep neural networks
• Author: Kristian Kapetanović
• Mentor: Ivan Marković (mentor)
• Committee member: Ivan Marković (predsjednik povjerenstva)
• Committee member: Marija Seder (član povjerenstva)
• Committee member: Ivan Petrović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-09-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering

Abstract

Mnoge primijene u robotici zahtijevaju gustu estimaciju dubine iz stereo slika. Iz toga je razloga potrebno generirati točne mape dispariteta u stvarnom vremenu, često i na ograničenim računalnim resursima. Klasične su metode relativno uspješne u tom zadatku, ali mogu biti računski vrlo složene. U posljednje se vrijeme analizira primjena dubokih neuronskih mreža za problem estimacije dubine jer one mogu iskoristiti različite izvore informacija u slici i postići kratko vrijeme izračuna na specijaliziranim računalnim arhitekturama. U radu su obrađene umjetne neuronske mreže, njihovo treniranje te su analizirane različite postojeće arhitekture.

Abstract (english)

Many applications in robotics require dense depth estimation from stereo images. For that very reason, it is necessary to generate correct disparity maps in real-time, often with limited computing resources. Classic methods are relatively successful but they can be quite complex for computation. Lately, many researches have been conducted regarding the problem of depth estimation because they can use various sources of information from images and achieve short computation times using specialized computational architectures. The thesis deals with the topic of artifical neural networks, the process of training them and analyzes various existing architectures

• Keywords: neuronske mreže; duboko učenje; stereo slike; disparitet; estimacije dubine
• Keywords (english): neural networks; deep learning; stereo images; disparity; depth estimation
• Language: croatian
• URN:NBN: urn:nbn:hr:168:471938
• Study programme: Title: 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 resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 12:38:08