Detection of Traffic Sign Based on Improved YOLOv4

To improve safe driving and reduce the mental fatigue of drivers, accurate and fast access to traffic sign guidance information in road scenes is needed. This work proposes a traffic sign detection method based on improved YOLOv4. First, the MLP architecture is combined with a convolutional neural network to enhance the network's ability to detect objects. Second, the same-level connection is added within PANet to enhance the feature information extracted by the network. Experiments show that the improved YOLOv4 algorithm can improve the detection accuracy of 416×416 sample images from 72.95% to 78.84%, and still meet the real-time detection requirements. After comparing our method with similar algorithms, the experimental results show that our algorithm has certain advantages.