Traffic Sign Detection Algorithm based on improved YOLOv4

Abstract

In the deep learning algorithm, YOLOv4 and Faster R-CNN have achieved excellent target detection performance. To improve the real-time detection of small targets in traffic signs, this paper presents an improved YOLOv4 target detection algorithm, we use K-means clustering to design anchor box, which is used to adapt to different small and medium scale. According to the size of small and medium scale signs, one more feature layer is extracted, and four different feature layers are fused to detect. Experimental results show that the improved algorithm can improve the detection accuracy and real-time performance.