Application of Chinese Traffic Sign Detection Based on Yolov4

With the social progress and the continuous development of science and technology, people have made great progress in the field of automatic driving and assisted driving. In automatic driving system, it is very important to achieve higher precision and faster detection and recognition for traffic sign. In this paper, an improved algorithm TSnet based on Yolov4 is proposed to realize the application of Chinese traffic sign detection in visual direction. We further optimize the detection performance of Yolov4 in this scene, restructure and adjust the network structure of Yolov4, use the idea of DenseNet to replace the residual unit with dense connection unit in the backbone network, so as to better extract features. Experiments show that our proposed network is superior to the original Yolov4 network, it can achieve the task of Chinese traffic sign detection better.