EVCW-YOLO: An Enhanced YOLOv11-Based Network for Electric Vehicle Helmet-Wearing Detection

To address the issues of small target recognition difficulties, insufficient feature representation ability, and limited positioning accuracy in electric vehicle helmet wearing detection under complex road environments, this paper proposes an efficient detection network called EVCW-YOLO based on the improved YOLOv11 network. First, introduce the EfficientViM and CGLU modules into the backbone network to optimize the existing C3K2 structure and enhance the model's feature extraction capability. Second, design a Multi-Scale Feature Enhancement Aggregation (MSEA) module to improve the model's detection performance for targets of different scales, especially small targets. Finally, incorporate a Weighted IoU (WIoU) loss function to optimize object localization effects and accelerate model convergence. Experiments were conducted on an electric vehicle helmet detection dataset. The results demonstrate that, compared to the baseline model, the proposed method improves mAP@ 0.5 by 2.5 %, with precision and recall increasing by 1.4 % and 3.4 %, respectively. These results indicate that the proposed model offers enhanced capability in detecting helmets under road conditions and exhibits promising potential for practical applications.