Abstract As electric automobiles (EVs) become more popular for safety, helmet laws must be enforced. Traditional detection approaches often lack real-time performance and flexibility in complex situations, thereby limiting their practicality. This research uses the Swin Transformer architecture and deep learning to create a real-time helmet-wearing detection system. The Swin Transformer performs well in visually demanding settings, including scenes with shifting lighting, complex backdrops, small item sizes, and partial occlusions, because it can model distant connections and maintain hierarchical representations. This research utilized a publicly available dataset of 10,000 photos of electric car passengers in various poses, perspectives, and scenarios. Modern object recognition frameworks, including Faster R-CNN, YOLOv5, DETR, and simple deep learning models, were used to evaluate the proposed model. Experimental findings show that the recommended technique performs better with a 95.3% F1-score, 94.5% recall, and 96.2% accuracy. The findings demonstrate significant improvements in accuracy and robustness over previous methods. This research presents a reliable real-world technique for identifying electric vehicle helmets. Integrating the suggested technologies with advanced traffic monitoring systems would assist public safety and law enforcement. This approach enhances safety compliance by leveraging cutting-edge deep learning for intelligent transportation.
Zhao et al. (Wed,) studied this question.