Recognising license plate from image data using deep learning

Abstract

Maintaining surveillance and security on the roads has become a significant challenge, which is why it has become necessary to conduct proper methods to control this problem. To ensure security and safety on the roads, methods such as Automatic Number Plate Recognition is implemented so that all crimes and other security-related issues may scale down. In this paper, two versions of YOLO are used. The first one is YOLOv5, and afterward, the most recent model of YOLO, which is YOLOv7. These models are used so that we may get better results com- pared to all other previously used models. Eventually, EasyOCR is used to extract the characters from the number plate. The proposed models are tested on the LP dataset, our custom dataset consisting of 10,700 images. 93.8% and 95.6% accuracy are acquired from YOLOv5 and YOLOv7, respectively. However, the main goal of this research paper is to prove that YOLO is superior to other models in terms of object detection. In addition, YOLOv7 provides us with improved results compared to YOLOv5.