AI-Powered Smart Helmet Detection System for Traffic Safety

Road accidents involving two-wheeler riders remain a critical global issue, with a substantial proportion of fatalities attributed to non-compliance with helmet-wearing regulations. Conventional enforcement approaches are often labor-intensive, prone to errors, and difficult to scale. To address these challenges, this study introduces an AI-powered helmet detection system that leverages real-time image and video data from traffic cameras and drones to automatically identify helmet violations. The proposed system employs Hybrid Attention-YOLO (HAYOLO), an advanced deep learning model that enhances the YOLO (You Only Look Once) framework by incorporating a dual-attention mechanism along with adaptive feature fusion, significantly improving detection accuracy under challenging conditions such as low-light environments, occlusions, and varying helmet designs. This system offers a scalable, efficient, and automated solution to enhance road safety and regulatory compliance.