Traffic Sign Classification using Transformers

Traffic sign classification plays a crucial role in intelligent transportation systems (ITS) by ensuring accurate sign recognition for autonomous driving and Advanced Driver Assistance Systems (ADAS). However, traditional models struggle with challenges such as varying lighting conditions, occlusions, and visually similar signs, leading to misclassifications. This study presents an enhanced YOLOv5 model integrated with Channel Attention (CA) to improve feature extraction and detection accuracy. The proposed approach dynamically prioritizes essential features, suppressing background noise and improving classification in complex environments. The model is trained and evaluated on a diverse traffic sign dataset, and performance is analyzed using key metrics such as accuracy, precision, recall, and mean average precision (mAP). Results indicate that YOLOv5 + CA significantly outperforms standalone YOLOv5, achieving higher recall, F1-score, and classification accuracy, making it a robust solution for real-time traffic sign recognition.