Traffic Sign Detection Based on Improved YOLOv5 Model

Traffic sign detection is an important issue in the fields of automatic driving. Aiming at the problem of YOLOv5 algorithm in detection small target traffic signs, an algorithm based on improved YOLOv5 is proposed. First, the Coordinate Attention (CA) model was integrated into the backbone back-end to capture the relationship between location information and channels, so as to obtain the region of interest more accurately. Then, the upsampling method in the feature fusion network was changed to transposed convolution to improve the learning ability of the network and the detection accuracy. Finally, cross-layer connections were introduced in the feature fusion network to fuse more shallow feature information, thereby improving the feature extraction ability of the network and further improving the detection accuracy of small objects. The experimental results show that the improved algorithm achieves 89.73% mAP on the traffic sign detection data set TT100K, which is 7.1% higher than the original YOLOv5. Therefore, the proposed improved YOLOv5 algorithm can effectively improve the accuracy of traffic sign detection.