Real-time Automatic License Plate Recognition System using YOLOv4

We introduce a real-time Automatic License Plate Recognition system that is computationally lighter by eliminating the ROI setting step, without deteriorating recognition performance. Conventional license plate recognition systems exhibit two main problems. First, clear license plate visibility is required. Second, processing actual field data is computationally intensive and the ROI needs to be set. To overcome these problems, we performed plate localization directly on the entire image, and conducted research taking low quality license plate detection into account.We aim to recognize the license plates of cars moving at high speeds on the road as well as stationary cars using the NVIDIA Jetson TX2 module, which is an embedded computing device.