Traffic-Sign-Detection Algorithm Based on SK-EVC-YOLO

Traffic sign detection is an important research direction in the process of intelligent transportation Internet era, and plays a crucial role ensuring traffic safety. The purpose this to propose traffic-sign-detection algorithm based on selective kernel attention (SK attention), explicit visual center (EVC), YOLOv5 model address problems small targets, incomplete detection, insufficient accuracy natural complex road situations. First, feature map with smaller receptive field backbone network fused other scale maps increase target layer. Then, SK mechanism introduced extract weigh features at different scales levels, enhancing target. By fusing gather local area within layer, effect targets improved. According experiment results, mean average precision (mAP) Tsinghua-Tencent Sign Dataset (TT100K) for proposed 88.5%, which 4.6% higher than original model, demonstrating practicality signs.