Traffic sign recognition based on deep learning

Abstract Intelligent Transportation System (ITS), including unmanned vehicles, has been gradually matured despite on road. How to eliminate the interference due various environmental factors, carry out accurate and efficient traffic sign detection recognition, is a key technical problem. However, traditional visual object recognition mainly relies feature extraction, e.g., color edge, which limitations. Convolutional neural network (CNN) was designed for based deep learning, successfully overcome shortcomings of conventional recognition. In this paper, we implement an experiment evaluate performance latest version YOLOv5 our dataset Traffic Sign Recognition (TSR), unfolds how model in learning suitable TSR through comprehensive comparison with SSD (i.e., single shot multibox detector) as objective paper. The experiments project utilize own dataset. Pertaining experimental results, achieves 97.70% terms [email protected] all classes, obtains 90.14% mAP same term. Meanwhile, regarding speed, also outperforms SSD.