A Novel Real-Time Helmet Wearing Detection Technique of Motorcyclists Using Fine-Tuned YOLOv8 Model for Indian Urban Road Traffic

Motorcycle is a rapidly growing industry. But it has maximum fatal rate of accidents than heavy vehicles. Above 37 million of people in India use motorcycle for transportation. Out of all the road accidents happening in India approximately 45% of accidents are caused by motorcycles and among them 32% of accidents are caused due to not wearing helmets. With the heavy traffic condition, the road transport authority essentially require automation model to detect wearing helmet of two wheelers and penalize if any violation of laws. This paper addresses such problem of heavy Indian urban road traffic conditions and proposed a fine-tuned deep learning based YOLOv8-SS2 helmet wearing detection technique. This model classifies helmet and non-helmet riders. The proposed model outperforms while comparing with existing literature having high accuracy of 95% and Mean Average Precision (m-AP) of 99%.