Energy and Cost Efficient Pre-trained Convolutional Neural Network (ResNet-50) Model for Tomato Disease Recognition Using Cloud Platform

Abstract

Tomato (Solanum Lycopersicum) is a popular crop worldwide. The harvesting of tomatoes is highly profitable for farmers but due to some diseases in a tomato plant, farmers suffer from several problems. Due to these diseases, the overall production of tomato fruits is reduced. The main goal of this research is to raise the overall production of tomato fruits. To accomplish this task, we have to identify the infection in tomato leaves at its initial stage and work to make the tomato plant infection-free. To recognize the disease at its initial stage, we have used an image classification model to resolve such types of problems. We have used a CNN model with residual network-50 as a learning algorithm for tomato disease classification. With the help of this algorithm, we have classified the infected tomato leaves and healthy tomato leaves separately by using MATLAB 2019(a). We have collected 6594 images of tomato leaves for training and evaluation. We have taken fifty percent of the entire data set for training & fifty percent of the dataset for testing purposes for ResNet-50. These leaves are contaminated by six different tomato diseases. After applying the above model, we obtained an effective outcome with 99.83% accuracy in a time span of 47.68 min only. The novelty of this research paper is to calculate time with accuracy. This model can be useful for farmers to defend tomato plants in opposition to disease. Our experiments exhibit the use of a pre-trained model for disease classification with better accuracy.