Publication

Abstract : The model is developed using YOLO v8 for detecting Motorbike riders who are violating road rules. Initially, the model differentiates motorbike riders who are wearing safety bike helmets from those who are not wearing them. Riders without helmets are marked and processed by OCR to detect number-plates. Images with helmets are checked if used with cameras attached to them. If detected, it is marked and processed by OCR to detect number-plates. Detected number-plates are classified under offense. The whole model is run on a custom dataset that we prepared which contains real-life road images. Then the images that are found to violating rules will be loaded to a OCR for number-plate detection. Bounding boxes is used to represent the detected objects and has made a prediction of 98% accuracy. The numbers detected from the OCR will be exported into a csv file and any duplicates removed.