Identification of Tomato Pests and Diseases Based on Transfer Learning

Abstract

There are numerous kinds of tomato diseases and insect pests. Their pathology is complex and different. It is hard to rely on manual identification purely and the error rate is high. After collecting a mass of leaf table pictures, our aim is to classify nine kinds of common tomato diseases in China. The idea of transfer learning is applied to achieve recognition and classification of tomato data set by the lightweight convolutional neural MobileNet. Finally, the model can obtain test classification accuracy of 97.19%. Experiments have proved that this method is not only simple to operate and easy to implement, but also can achieve high accuracy on plant diseases.