A Comparative Study of YOLO and Deep Neural Network Models for Vehicle Number Plate Detection

View at publisher Summarise with VIRAT
Authors: Ebai Geslen Muwa, Vaneet Kumar, Ruchika Gupta
Cited by: 0
References: 3

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Due to the increasing number of commercial vehicles, fluctuating weather conditions, and long-existing cars with extremely noisy back-grounds, Automatic Number Plate Recognition (ANPR) is crucial in improving traffic management, optimizing transportation networks, and enhancing security zone access control. The accuracy of different license plate detection models trained on diverse datasets was analysed, compared, and evaluated to propose an improved system for the automatic license plate detection and recognition of characters. The objective was to achieve state-of-the-art license plate detection and character recognition accuracy, particularly in challenging scenarios. Several deep learning models were explicitly trained on manually collected and annotated datasets under various environmental conditions. It was shown that YOLOv5 (You Only Look Once model) outperformed other models, achieving a consistent average of 99.69% for detection and 98.37% for recognition across multiple conditions, including rainy, cloudy, and nighttime scenarios.

Fields of Study & Subjects

Vehicle License Plate Recognition Advanced Neural Network Applications Autonomous Vehicle Technology and Safety

Publication details

Journal / source
International Conference on Signal Processing and Electronics Design (ICSPED)
Pages
77–80
Publication date
23 January 2026
Language
EN