Assessing the Accuracy of Object Detection and Image to Text Conversion on Motor Vehicle Number Plate Using YOLOv5 and PaddleOCR

Object detection and image conversion to text is a widely used technology, with various applications, including the detection of Motor Vehicle Number Plate (TNKB) and conversion of the image to text. This research aims to understand how to create a TNKB detection and image conversion system and measure accuracy in the process of object detection and image to text conversion in TNKB. The research uses the YOLOv5 algorithm for object detection and the PaddleOCR library to convert TNKB images into text. The development method that is being used are CRISP-DM. This research involved the use of 298 images of four- wheeled motorized vehicles obtained from various sources, including Google Image Search, Instagram, and the government auction website lelang.go.id This data is used to train the YOLOv5 model. In addition, 40 images of four-wheeled motorized vehicles from the same source were used for testing. The research results show that the YOLOv5 algorithm achieves detection accuracy of 95%, while PaddleOCR's accuracy in converting TNKB images into text reaches 75%.