Health Detection for Potato Leaf with Convolutional Neural Network

Potato is the fourth largest food crop in the world and grown in many places of the world. Potato crops mainly infected with fungi, and hence they got early blight diseases and late blight diseases. Real time control of disease and management can effectively increase production and reduce farmers' losses. The ability can identify infected crops automatically for farmers. Therefore, this paper proposes a CNN (Convolutional Neural Network) architecture which is suitable for potato disease detection. At first, we will create a database for our training set by means of image processing in the CNN. Adam is used as the optimizer, and cross entropy is used as the model analysis basis. Softmax is used as the final judgment function. The convolution layer and resources are minimized usage amount while maintaining high accuracy. The experimental results show that the parameter usage is 10,089,219 and the accuracy of the disease judgment can reach 99% under the preset model which is proposed in this paper.