YOLOv8 for Bangla License Plate Recognition: Advancing Real-Time Object Detection in Localized Contexts
Utsha Saha, Imam Uddin Ahamed, Mohammed Hossain
Abstract
Automatic License Plate Recognition (ALPR) systems play a crucial role in traffic management and security applications, and the development of accurate and efficient ALPR models is essential for their success. This study investigates the application of YOLOv8, a state-of-the-art object detection architecture, for recognizing Bangladeshi license plates. A comprehensive dataset, combining images from various vehicle categories and environments, was curated and augmented to train five variants of the YOLOv8 model: YOLOv8n, YOLOv8s, YOLOv8m, YOLOv81, and YOLOv8x. The models were rigorously evaluated using metrics such as mean Average Precision (mAP), precision, recall, and F 1 -score. The results demonstrate the effectiveness of YOLOv8 in detecting and recognizing Bangladeshi license plates, with YOLOv8x achieving the highest mAP50 and mAP5095 scores of 0.96 and 0.75, respectively. The study highlights the adaptability of YOLOv8 to the unique characteristics of Bangladeshi license plates and provides insights into model selection criteria based on computational resources and performance requirements. Future research directions include expanding the dataset to cover a wider range of environments and integrating the YOLOv8 models into a comprehensive ALPR system for enhanced traffic management and security applications.