Tomato leaf disease identification based on improved AlexNet network
In response to the issues of extensive training parameters and prolonged training time encountered by tradi-tional convolutional neural networks in identifying tomato leaf diseases,an optimized network based on AlexNet,termed Feature Reuse Network(FRNet)is proposed and experimentally validated.The experimental results showed that compared to the conventional AlexNet model,the FRNet significantly reduces both training time and model size.The parameter count of the FRNet model was only 1.12%of that in AlexNet,and the model size was also significantly reduced to 2.51 MB,while achieving an average recognition accuracy of 98.82%.Compared to other methods,FRNet not only exhibited higher recogni-tion accuracy but also proved to be suitable for developing lightweight mobile based systems for tomato leaf disease recogni-tion,providing a new solution for agricultural disease identification.
NETL
NSTL
万方数据
维普
