A Novel Apple Leaf Disease Recognition Method Based on Improved ResNet18 Neural Network
In order to effectively improve the accuracy and efficiency of apple leaf disease recognition,and achieve timely prevention and treatment of apple diseases so as to improve yield,this study proposed an apple leaf disease recognition method based on the improved ResNet18 neural network,which could improved the recognition performance of the model but reduce the parameter quantity and model size.First,improved the residual structure of ResNet network to reduce the parameter quantity,which could achieve the model lightweighting.Second,integrated the coordinate attention(CA)mechanism and transfer learning into the model to further improve its generalization performance.Comparing with the original ResNet18 model,the ac-curacy of the improved model increased by 1.53 percentage points,but the parameter quantity reduced to 50.84%of the original model.The above results indicated that the improved model could effectively recognize apple leaf diseases and was easy to carry on mobile terminal.
万方数据
维普
