Razvoj aplikacije za detekciju i raspoznavanje prometne signalizacije iz video signala korištenjem modela dubokog učenja

undergraduate thesis

• Title (english): Development of an application for detection and recognition of traffic signs from a video signal using deep learning models
• Author: Jure Franjković
• Mentor: Marko Horvat (mentor)
• Committee member: Marko Horvat (predsjednik povjerenstva)
• Committee member: Juraj Dončević (član povjerenstva)
• Committee member: Kristina Ćaleta (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-08, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom radu korišteni su modeli dubokog učenja, konkretno YOLOv8, za detekciju prometne signalizacije. Primijenjeno je znanje iz umjetne inteligencije, dubokog učenja i računalnog vida. Model je treniran na skupu slika hrvatskih prometnih znakova koristeći platformu Kaggle, koja omogućuje upotrebu naprednih grafičkih kartica za brže treniranje. Evaluirane su različite metrike evaluacije detektora objekata, uključujući box_loss, cls_loss, dfl_loss, preciznost i opoziv. Rezultati su pokazali da model postiže visoku točnost, iako postoje izazovi u detekciji semafora. Dobiveni model je implementiran u aplikaciju koja prima slike i video signale i detektira objekte u njima. Projekt pokazuje potencijal za poboljšanje sigurnosti prometa kroz napredne tehnologije.

Abstract (english)

In this study, deep learning models, specifically YOLOv8, were used for the detection of traffic signals. Knowledge from artificial intelligence, deep learning, and computer vision was applied. The model was trained on a dataset of Croatian traffic sign images using the Kaggle platform, which enables the use of advanced graphics cards for faster training. Various object detection evaluation metrics were evaluated, including box_loss, cls_loss, dfl_loss, precision, and recall. The results showed that the model achieves high accuracy, although there are challenges in detecting traffic lights. The obtained model was implemented in an application that receives images and video signals and detects objects in them. The project demonstrates the potential for improving traffic safety through advanced technologies.

• Keywords: Duboko učenje; YOLOv8; prometna signalizacija; detekcija objekata; Kaggle; metrike evaluacije detektora objekata
• Keywords (english): deep learning; YOLOv8; traffic signals; object detection; Kaggle; object detection evaluation metrics
• Language: croatian
• URN:NBN: urn:nbn:hr:168:715808
• 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: Open access
• Created on: 2024-10-12 20:35:14