Traffic Sign Recognition in Rainy, Snowy and Foggy Weather Based on Improved YOLOv8
Cong Zhang · 2024
To address the issue of poor performance in traffic sign recognition technology during adverse weather conditions like rain, snow, and fog, this paper presents an enhanced YOLOv8 model. This model excels in enhancing the accuracy of traffic sign recognition, significantly outperforming the original YOLOv8 in terms of performance. We optimized the model's Neck section by substituting the original PANet with a Weighted Bidirectional Feature Pyramid Network (BiFPN), thereby bolstering the model's feature fusion capabilities. Additionally, we incorporated a Global Attention Mechanism (GAM) into the Neck region to precisely adjust attention during the crucial feature fusion phase, further elevating detection accuracy. Compared to the original YOLOv8 model, our improved model exhibits a 4% increase in the mean Average Precision (mAP) score, facilitating more precise detection of traffic signs.