IR-LPR: Large Scale of Iranian License Plate Recognition Dataset

TL;DR

This paper introduces IR-LPR, a comprehensive large-scale dataset of Iranian license plates with over 20,000 images for detection and nearly 28,000 images for character recognition, addressing the scarcity of such datasets.

Contribution

The creation of the largest Iranian license plate dataset with detailed annotations for detection and recognition tasks, facilitating research and commercial applications.

Findings

Provides a comprehensive dataset for license plate detection and recognition.

Enables development of deep learning models for Iranian license plates.

Fills a significant gap in available datasets for Iranian license plate recognition.

Abstract

Object detection has always been practical. There are so many things in our world that recognizing them can not only increase our automatic knowledge of the surroundings, but can also be lucrative for those interested in starting a new business. One of these attractive objects is the license plate (LP). In addition to the security uses that license plate detection can have, it can also be used to create creative businesses. With the development of object detection methods based on deep learning models, an appropriate and comprehensive dataset becomes doubly important. But due to the frequent commercial use of license plate datasets, there are limited datasets not only in Iran but also in the world. The largest Iranian dataset for detection license plates has 1,466 images. Also, the largest Iranian dataset for recognizing the characters of a license plate has 5,000 images. We have prepared a complete dataset including 20,967 car images along with all the detection annotation of the whole license plate and its characters, which can be useful for various purposes. Also, the total number of license plate images for character recognition application is 27,745 images.