Research on Road Disease Identification Based on YOLOX-tf2-EffcientNet Lightweight Model
In order to enhance the efficiency of utilizing three-dimensional ground-penetrating radar(GPR)images,this study proposes the YOLOX-tf2-EfficientNet network framework.Utilizing image analysis and data augmentation techniques,an internal defect dataset for three-dimensional GPR is created.The high-integration computing capability of GPU is fully leveraged to input this dataset into the YOLOX-tf2-EfficientNet network architecture.During testing,the model achieves an average detection accuracy of 75.59%,meeting the accuracy requirements in engineering applications.In terms of dataset testing,this approach effectively detects various issues within the road and accurately locates them.
万方数据
维普
