A Machine Learning Approach to Pomegranate Leaf Disease Identification

Abstract

Pomegranate is a fruit with the highest yield and the greatest geographic distribution in Asia. On the other hand, the plants are susceptible to a diverse range of illnesses as a consequence of a number of factors, which leads to the total destruction of the plants and a harvest that is significantly reduced. In order to avoid reductions in agricultural output, it is necessary to diagnose plant diseases as quickly as is practically possible. It is a difficult and time-consuming task to manually monitor the progress of diseases on pomegranate leaves. Therefore, Deep Learning is utilised in the diagnosis of diseases affecting pomegranate trees (DL). The goal of this study is to develop an algorithm for automatically diagnosing diseases affecting pomegranate plants based on images of the plant's leaf structures. The process of a disease detection system includes the gathering of data, the analysis of that data, the categorization of gathered images, and their subsequent deployment. The Mendeley database is utilised in order to generate images of pomegranate leaves in both healthy and diseased states. After that, the original, unaltered raw image is polished. Two different DL models, AlexNet and VGG-16, are put to use in this classification technique. To determine the optimal model, it is necessary to do accurate and loss-oriented measurements. According to the measurements, AlexNet does a good job of recognising diseases that affect leaf tissue. Later, an AlexNet-based smartphone app is developed to assist farmers in performing disease detection on pomegranates without the assistance of professionals. This software is intended to help farmers save time and money.