LightWeight Helmet Wearing Detection Algorithm for Improved YOLOv8n

To address the issues of large model parameters, missed detections, false detections of small objects in complex backgrounds, and low detection accuracy in existing helmet wearing detection algorithms, an improved lightweight YOLOv8n-based helmet wearing detection algorithm named YOLOv8-RLL is proposed. A C2f-RVB module, integrating RepViTBlock, is designed to replace the original C2f module to reduce model parameters and achieve lightweight performance. Simultaneously, The LSKA attention mechanism is introduced in SPPF to ignore irrelevant background information interference and focus on more effective helmet features. A shared lightweight convolutional detection head (LSCD-Head) is implemented to improve the expressive power of multi-scale features and further reduce parameter quantity. Experimental results show that compared with the YOLOv8n model, the YOLOv8-RLL algorithm reduces parameters by 36.3%, decreases GFLOPS by 37%, reduces weight size by 34.9%, and increases mAP@ 0.5 by 1.2%, reaching 93.8%. This algorithm achieves lightweight safe helmet detection and improves average precision.