Semi supervised identification of surface cracks in stone carvings based on improved VGG16 network
In order to quickly and accurately detect the surface cracks in stone carvings,the authors establish an improved VGG16 network model,which changes the original three fully connected layers to two fully connected layers,adds dropout regularization,and combines it with the semi supervised learning algorithms to apply deep learning to intelligent recognition research of surface cracks in stone carvings.In order to test the performance and accuracy of the improved model,Unet,ResNet and the original models were selected for comparison.The improved VGG16 network model achieved an accuracy of 93.6％and the training time is reduced by 18％compared with the original models,which has the advantage of lightweight operation.The model can meet the basic needs of surface crack recognition and has good robustness.
stone carvingssurface crack identificationsemi supervised algorithmimproved VGG16 network
NETL
NSTL
万方数据
