YOLOv8 Safety Helmet Detection Algorithm Based on Attention Mechanism and Feature Fusion

Abstract Existing safety helmet detection algorithms face challenges such as insufficient feature extraction for small targets, poor adaptation to multi-scale targets and limited localization accuracy in architectural scenes. To solve the above problems, an improved YOLOv8 model is proposed in this paper. First, the GC block with integrated CBAM attention mechanism is introduced into the C2f module of the backbone network to significantly enhance the focusing ability in critical regions. Second, Ghost-Bottleneck modules with different scales are introduced into the neck network to extract features with different scales and provide rich feature information for the model; it significantly enhances the model’s ability to detect helmet targets with different sizes. Finally, the Wise-IoU loss function is introduced and an adaptive scale-aware mechanism is added to improve the target localization accuracy and robustness of the model in complex scenes. The experimental results on the SWHD dataset show that the improved model achieves 90.4% detection accuracy, which is 4.3% higher than the original model, and 6.2% higher than the mAP, showing significant advantages in complex scenes.