License Plate Detection and Recognition Using Deep Learning in Unconstrained Scenarios

License Plate Recognition was widely studied due to its extensive applications. Multi scenarios limitation and low recognition accuracy are two inherent problems of LPR. In this paper, an end-to-end LDPR network is proposed to address the former problem. We divide the recognition problem into Plate Detection and text recognition process. By using YoLov4 detection network the influence undertaken by scenarios change has been solved. In addition, a Convolutional Block Attention Module is employed in the YoLov4 Network to improve the recognition accuracy. A SRN based network is then added in to the network to improve the recognition rate of the LDPR systems. Experiment results demonstrate that the proposed algorithm outperforms most of the existing plate recognition algorithms.