License Plate Detection Using Deep Learning Object Detection Models

Object detection is an extension of image classification tasks in computer vision. Its goal is to locate an object of interest in any given image input. In the past, this was done by traditional hand-crafted feature algorithms i.e., SIFT, SURF, HOG, BRIEF, and ORB. These algorithms have been successful in their field however they do possess some downsides due to their nature. For example, they can be slow in detection speed, not as accurate, and difficult to develop. Since 2012, deep learning has become an emerging technology that can solve object detection problems with relatively better performance. However, not many studies have been done to deploy deep learning object detection models in real-world scenarios, e.g., license plate detection. License plate detection is a challenging task in computer vision because the input image captured can be in different sizes, colors, distances, orientations, and lighting conditions. This project aims to study and improve license plate detection using deep learning models. As of the current year, the model YOLOv4 has achieved 43.5% Average Precision (AP) on MS COCO. Meanwhile, EfficientDet-D7 has achieved 55.1 AP on COCO test-dev. In this paper off-the-shelves object detection models are trained on CCPD license plate dataset. Two approaches have been carried out to improve their accuracy, i.e., image preprocessing step, and modifying existing model architecture. Preprocessing steps show improvement for all test sets in terms of (TP-FN-FP) value, i.e., (69-31-45 to 75-25-43) for db test set, (70-30-28 to 73-27-28) for blur test set, (52-48-75 to 66-34-68) for fn test set, (92-8-9 to 97-3-4) for rotate test set, (77-23-26 to 85-15-18) for tilt test set, and (67-33-61 to 76-24-56) for challenge test set. The improved model has achieved (96.96%) mean Average Precision mAP@0.70 on the validation dataset compared to the original model (83.64%)