A New Benchmark Dataset for Egyptian License Plate Detection and Recognition

Automatic license plate detection and recognition have become an integral part of the new ecosystems to ensure safety and traffic management. It is also challenging because it depends on several factors, including the country’s design, colors, language, and texture. One of the main issues that face the deep learning models is the lack of resources and datasets, especially in the Arabic language. In this work, we propose a large-scale dataset, namely Egyptian Automatic License Plate Recognition (EALPR), to address this issue. The images are collected from social-media websites using scraping tools. After cleaning and sanity checks, the YOLO detector is used for annotation. The total number of vehicles is 2,450, including different locations, times (night and day), and backgrounds to guarantee the balance of the dataset. We also provide the detailed steps of the dataset collection and annotation. Finally, we discuss the comparison of the state-of-the-art open-source datasets. The dataset and models will be publicly available on https://github.com/ahmedramadan96/EALPR