A Diagnostic Method for Detecting Tomato Leaf Diseases Based on a Deep Learning Approach

Abstract

Nowadays, the agricultural area has been enhancing every day due to the innovation that makes farming far more effective and efficient in the production of crops. The Industrial Revolution 4.0 (IR4.0) plays a major role in ensuring that Smart Farming concept can be realized. For instance, a plant diagnostic tool using a deep learning approach can be developed to monitor the growth of plant or to investigate the health of the plant based on its leaf conditions. This study aims to investigate the feasibility of optimizing the deep learning algorithm to diagnose the quality of tomato crops using an optimized Convolutional Neural Network (CNN) model. In other words, this paper investigates the effects of varying the parameter settings used in the CNN architecture on the prediction accuracy of the tomato diseases. The objective of this project is to formulate a deep learning model, specifically a CNN model, that can be used to learn tomato leaf images to detect tomato diseases and also to optimize the proposed CNN model by utilizing the CNN parameters such as number of epochs, batch size and learning rate. The methodology used in this work can be divided into three main tasks which are Data Capturing and Fusion, Data Modelling and Model Assessment. Based on the obtained results, the combination of all three parameters has a great influence on the performance accuracy of the proposed deep learning algorithm. For instance, the highest performance accuracy of the CNN model could be obtained by setting the parameters as follows; Epoch = 500, Batch Size = 16, Learning Rate = 1e − 1, accuracy of 95.54% for the training dataset and 65.0% for the test dataset.