Deep Convolutional Neural Networks for image based tomato leaf disease detection

Plant leaf diseases are a major significant risk to food security. In many situation the agriculture production may be reduced, which consequently reduces the nation's economy, if the crops get affected due to diseases. Generally, diseases affect the leaves of the crops which should be identi-fied in the early stage so that the quality and quantity of the produce may be increased. To detect the leaf diseases at an early stage and taking proper remedial actions will be more helpful for the farmers. So there is a need for an automatic system for leaf disease recognition that identifies and classifies the leaf diseases at an early stage. The highlights of the objective work focus on the DCNN models of leaf images used and overall performance according to the performance metrics that are been applied for plant disease identification. During the past decade many researchers are focusing on leaf disease recognition by proposing various methods and techniques using tra-ditional image processing and machine learning techniques. This motivation of the proposed DCNN work is suited for increasing performance accuracy and minimizing response time in the identification of tomato leaf diseases. In this paper we have proposed an automatic system for leaf disease identification in tomato leaves using Deep Convolutional Neural Network (CNN) since, DCNN focuses on agriculture areas during recent years. In the proposed work we have used 18160 images of tomato leaf diseases which are collected from plant village data set. We have split the dataset that contains 60% of images from the dataset for training and 40% of images for testing. With our proposed DCNN model we have obtained 98.40% of accuracy for the testing set.