Identification of Tomato Plant Diseases by Leaf Image Using Squeezenet Model

Abstract

One of the problems in the field of agriculture is regarding plant diseases. Plant diseases can cause a decrease in agricultural production. Therefore, early detection and diagnosis of plant diseases is very important. Plant diseases often appear on the leaves, and the characteristics of the affected leaves can be varied and difficult to distinguish. This makes it difficult to identify the disease automatically. Increased smartphone usage and advances in the field of computer vision through deep learning have made it possible to connect smartphones as a tool in diagnosing diseases. Plants used as case studies in this study were tomato plants. The method used is Convolutional Neural Network (CNN). This CNN method has many types based on the architecture it builds, one of which is squeezenet architecture. Squeezenet architecture can produce a model with a relatively small size, so that the model can be implemented on smartphone devices, server computing, and microcontroller devices. This research will focus on building a model based on squeezenet architecture to classify seven types of tomato plant diseases on the leaves including healthy leaves. The model is built using the help of Keras deep learning frameworks. The data used in this study is the image of tomato plant leaves obtained from the Vegetable Crops Research Institute (Balitsa) in Lembang, West Java. With 200 classes for each class. This study has successfully detected tomato plant disease through its leaf image automatically with an average accuracy of identification of 86.92%.