Attention-YOLOV4: a real-time and high-accurate traffic sign detection algorithm.

The technology of traffic sign detection is an important technology in autonomous driving. However due to the small size of traffic signs and the complicated background there are some problems in the practical application of traffic sign detection technology: the detection accuracy of one-stage traffic sign detection algorithm is low, the detection speed of two-stage traffic sign detection algorithm is slow. To achieve real-time and accuracy detection for traffic signs, we treat traffic sign detection as an end-to-end problem in this paper and proposed an improved one-stage traffic sign detection algorithm: Attention-YOLO V4. In order to achieve real-time and high-accurate detection of traffic signs so we analyzed the principle of YOLO V4 for small objects detection: in order to improve the ability of YOLO V4 backbone network extracting traffic sign features we decide to combine channel attention mechanism with residual block, in order to improve the ability of YOLO Head detect traffic signs we combine channel attention mechanism with YOLO head. We used TT100K dataset to evaluate Attention-YOLO V4 algorithm, compared with the existing methods, our method achieve real-time and accurately performance in complex backgrounds. [ABSTRACT FROM AUTHOR]