Abstract: In order to improve the speed and accuracy of the vehicle perception system in recognizing traffic signs, a traffic sign recognition method using the Yolov4 (You only look once version 4) deep learning framework was proposed. This method was compared with the single shot multi box detector (SSD) and Yolov3 (You only look once version 3) algorithms, which showed that parameters of the proposed algorithm model had increased significantly. The backbone feature extraction network and multi-scale output of Yolov4 were further adjusted by the algorithm. And a lightweight Yolov4 algorithm was proposed. Experimental results showed that the improved algorithm could effectively detect traffic signs, and had good real-time performance and applicability.