TY - GEN
T1 - YOLOv5 for Automatic License Plate Recognition in Smart Cities
AU - Raza, Abir
AU - Badidi, Elarbi
AU - Badidi, Basma
AU - Al Zahmi, Sarah
N1 - Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2023
Y1 - 2023
N2 - Recognition and extraction of license plate information from still images or videos are the basis of modern traffic and security systems. Automatic License Plate Recognition (ALPR) is transforming public safety and transportation in many ways. Such license plate recognition systems enable advanced solutions for toll roads, offer significant operational cost savings through automation, and even open up new market opportunities, such as license plate readers mounted on police vehicles. In this work, we implement license plate recognition based on convolutional neural networks to achieve high accuracy. You Only Look Once (YOLO) algorithm is employed for detection, while, an Optical Character Recognition (OCR) technique is used for text recognition. We use four variants of the YOLOv5 for license plate detection and the EasyOCR for license plate recognition. The results show that YOLOv5x (extra-large) achieves a Mean Average Precision (mAP) of nearly 82% in detecting vehicles and license plates in an image. Furthermore, the recognition of the letters on license plates is achieved with a confidence score of nearly 65%.
AB - Recognition and extraction of license plate information from still images or videos are the basis of modern traffic and security systems. Automatic License Plate Recognition (ALPR) is transforming public safety and transportation in many ways. Such license plate recognition systems enable advanced solutions for toll roads, offer significant operational cost savings through automation, and even open up new market opportunities, such as license plate readers mounted on police vehicles. In this work, we implement license plate recognition based on convolutional neural networks to achieve high accuracy. You Only Look Once (YOLO) algorithm is employed for detection, while, an Optical Character Recognition (OCR) technique is used for text recognition. We use four variants of the YOLOv5 for license plate detection and the EasyOCR for license plate recognition. The results show that YOLOv5x (extra-large) achieves a Mean Average Precision (mAP) of nearly 82% in detecting vehicles and license plates in an image. Furthermore, the recognition of the letters on license plates is achieved with a confidence score of nearly 65%.
KW - Convolutional neural networks
KW - Deep learning
KW - Object detection
KW - Object recognition
UR - http://www.scopus.com/inward/record.url?scp=85171561129&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85171561129&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-42317-8_8
DO - 10.1007/978-3-031-42317-8_8
M3 - Conference contribution
AN - SCOPUS:85171561129
SN - 9783031423161
T3 - Lecture Notes in Networks and Systems
SP - 93
EP - 105
BT - The 4th Joint International Conference on Deep Learning, Big Data and Blockchain (DBB 2023) -
A2 - Younas, Muhammad
A2 - Awan, Irfan
A2 - Benbernou, Salima
A2 - Petcu, Dana
PB - Springer Science and Business Media Deutschland GmbH
T2 - 4th Joint International Conference on Deep Learning, Big Data and Blockchain, DBB 2023
Y2 - 14 August 2023 through 16 August 2023
ER -
