Deep Learning for Tomato Diseases: Classification and Symptoms Visualization

Abstract

Several studies have invested in machine learning classifiers to protect plants from diseases by processing leaf images. Most of the proposed classifiers are trained and evaluated with small datasets, focusing on the extraction of hand-crafted features from image to classify the leaves. In this study, we have used a large dataset compared to the state-of-the art. Here, the dataset contains 14,828 images of tomato leaves infected with nine diseases. To train our classifier, we have introduced the Convolutional Neural Network (CNN) as a learning algorithm. One of the biggest advantages of CNN is the automatic extraction of features by processing directly the raw images. To analyze the proposed deep model, we have used visualization methods to understand symptoms and to localize disease regions in leaf. The obtained results are encouraging, reaching 99.18% of accuracy, which ourperforms dramatically shallow models, and they can be used as a practical tool for farmers to protect tomato against disease.