An Improved Diagnostic Approach for Classifying Tomato Leaf Diseases using Ensemble Deep Learning based Technique

Tomato is one of the most common vegetables that has high commercial importance for abundant production to meet the huge demand world-wide. There are several challenges that put impact on the mass production of tomato, and which are required to be resolved timely. One of the major challenges is tomato leaf diseases where different types of diseases are there which need to be identified correctly to protect the plants, otherwise it can cause severe harm to the plants which ultimately effects the production of tomato. To have a solution, several scientists and researchers have been working on this to classify diseases in a very initial stage. Diseases of the plants can be classified from leaf image, but it’s difficult to identify disease manually by observing through eyes. Deep learning techniques are being used to identify different types of disease using the leaf image. Several existing approaches have been discussed in this study where different detection processes have been proposed. But, in most of the cases, several limitations have been observed regarding the performance of those traditional frameworks. In this study, a novel disease detection approach has been proposed to detect multiple types of tomato leaf disease through analyzing the leaf images. Here, after combing the performances of four CNN models (MobileNetv2, ResNet50, VGG19, VGG16), an ensemble deep learning technique has been applied to identify different types of the tomato leaf disease as well as to classify the healthy and affected leaf. The performance achieved from the proposed ensemble framework ensures optimum accuracy of 96.7% with efficiency. A proper comparative analysis has been presented that supports the novelty and potentiality of the work.