Research on rice leaf disease identification based on ResNet

The traditional identification of rice leaf disease has a long period and low accuracy, which depends on artificial design features. In this paper, a rice leaf disease identification network based on depth residual network is proposed. The network is based on ResNet 101 network, and the nonlinear SVM algorithm based on kernel function is introduced to make the data samples linearly divisible. Secondly, the plant Village data set is migrated to the parameters trained in ResNet 101 network to complete the construction. After verification, the network can better balance the requirements of recognition accuracy and network lightweight and efficient, and the average recognition accuracy of the model is as high as 99.89%. By observing the evaluation criteria of the models such as accuracy rate, it can be seen that the network proposed in this paper has higher comprehensive average recognition rate, faster convergence speed, better robustness and generalization ability than the reference model in rice leaf disease identification, and has good application prospects, and initially meets the production requirements of rice disease identification.