Concrete Crack Detection Method Based on Improved YOLO Network
The target detection network with large models and high computing power requirements is difficult to im-plement mobile deployment,which restricts its application in the field of concrete crack detection.A concrete crack detection network model(YOLO-GAMM)is proposed,in which the backbone network of YOLOv5s is replaced by a lightweight network MobileNetV3 and a global attention mechanism is introduced to achieve YOLO-GAMM,so that its complexity and accuracy can satisfy the requirements of portable concrete crack detection systems.The sim-ulation results show that compared with the YOLOv5s model,the improved network model has a 60.0%reduction in computation and a 65.5%reduction in memory consumption and has a good performance.
国家科技期刊平台
NSTL
万方数据
