Detekcija objekata u stvarnom vremenu korištenjem dubokih neuronskih mreža

undergraduate thesis

• Title (english): Real-time object detection using deep neural networks
• Author: Katarina Jurič
• Mentor: Ivan Marković (mentor)
• Committee member: Ivan Marković (član povjerenstva)
• Committee member: Luka Petrović (član povjerenstva)
• Committee member: Karlo Koledić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-09-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Kroz rad je opisana povijest razvoja računalnog vida. Objašnjena je arhitektura
te princip rada konvolucijskih neuronskih mreža. Za detekciju objekata u stvarnom
vremenu korišten je YOLOv3 algoritam, a rad sarži njegov opis kao i usporedbu s dru-
gom sličnom konvolucijskom mrežom Fast R-CNN. Napisan je izvršni file koji u Ro-
botskom Operacijskom Sustavu pokreće web kameru i YOLO što rezultira detekcijom
objekata. Provjeravala se ovisnost točnosti detekcije objekta o njegovoj udaljenosti od
web kamere. Najvažniji zaključak je da detekcija ovisi o kontekstu i ostalim objektima
u slici.

Abstract (english)

The thesis describes the history of computer vision development. The architecture
and working principles of convolutional neural networks are explained. The YOLOv3
algorithm was used for object detection in real time and the paper summarizes its
description as well as a comparison with another similar convolutional network - Fast
R-CNN. An executable file has been written in which the Robot Operating System
starts the web camera and YOLO. The dependence of object detection accuracy on its
distance from the web camera was checked. The most important conclusion is that
detection depends on the context and other objects in the image.

• Keywords: računalni vid; konvolucijske neuronske mreže; YOLOv3; detekcija objekata; ROS
• Keywords (english): computer vision; convolutional neural networks; YOLOv3; object detec- tion; ROS
• Language: croatian
• URN:NBN: urn:nbn:hr:168:261663
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-15 10:26:42