Monokularna estimacija dubine u autonomnoj vožnji uz poznatu visinu kamere

master's thesis

• Title (english): Monocular depth estimation in autonomous driving with known camera height
• Author: Karlo Dinčir
• Mentor: Ivan Marković (mentor)
• Committee member: Ivan Marković (član 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: 2023-07-11, Croatia

Abstract

Estimacija dubine je jedna od najvažnijih informacija iz scene za veliki broj primjena i najbolji rezultati se ostvaruju se neuronskim mrežama i fuzijom senzora. Zbog male cijene i dostupnosti monokularne kamere pokušava se estimirati točna dubina samo monokularnom kamerom. Problem monokularne kamere je nemogućnost određivanje dubine koja ima metrički točnu skalu. Relativna dubinska mapa dobivena iz monokularne kamere u ovom radu će se pretvoriti u dubinsku mapu s metrički točnom skalom iz informacije o visini kamere. Kako bi izbjegli skupi i vremenski zahtjevan proces sakupljanja točno označenih podataka, koristit ćemo samonadziranu mrežu koja sama točno označava podatke za treniranje.

Abstract (english)

Depth estimation is one of the most important pieces of information from the scene for a wide range of applications, and the best results are achieved through neural networks and sensor fusion. Due to the low cost and availability of monocular cameras, an attempt is made to estimate accurate depth using only a monocular camera. The problem with monocular cameras is the inability to determine depth with a metrically accurate scale. The relative depth map obtained from a monocular camera in this study will be converted into a depth map with a metrically accurate scale using camera height information. To avoid the expensive and time-consuming process of collecting accurately labeled data, we will use a self-supervised network that self-labels the training data accurately.

• Keywords: estimacija dubine; monokularna kamera; konvolucijske neuronske mreže; samonadzirano učenje
• Keywords (english): depth estimation; monocular camera; convolutional neural network; self-supervised learning
• Language: croatian
• URN:NBN: urn:nbn:hr:168:906883
• Study programme: Title: Information and Communication Technology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije (magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-10-11 08:43:11