A YOLO11-based lightweight traffic sign detection model

This paper introduces a lightweight enhanced model based on YOLO11 to tackle the dual challenges of reducing model size and boosting precision for deploying traffic sign detection systems on autonomous driving mobile devices.Initially, the backbone network integrates the C3G2 module, which effectively minimises parameters while strengthening multi-scale feature extraction.Subsequently, a parameter-free attention mechanism, Tr-simAM, is developed to lower computational load and heighten sensitivity to small targets.Finally, the neck network incorporates the lightweight dynamic upsampling module DySample to enhance feature fusion and localisation precision.Experimental outcomes reveal that the refined algorithm achieves an 84% precision rate, signifying a 1.2% gain over the baseline YOLOv11n model, while model parameters and computational resources decrease by 19% and 7%, respectively.These lightweight enhancements efficiently fulfil practical requirements for real-time performance and computational constraints in autonomous driving mobile devices.