Development of a Pomegranate Fruit Disease Detection and Classification Model using Deep Learning

Abstract

Background: India’s economy heavily relies on agriculture, which provides a living for a significant portion of the nation’s population. Agriculture in India faces several challenges, including fruit diseases. Pomegranate cultivation is widespread across various states in India, with Maharashtra, Karnataka Andhra Pradesh, Gujarat and Tamil Nadu being major pomegranate-producing states. Fruit disease detection and classification play a significant part in refining crop yield: water scarcity, soil degradation, outdated farming techniques and the impact of climate change. Farmer suicides are a distressing issue, often linked to financial burdens, crop failures and debt. Manual disease detection methods are labor-intensive, time-consuming and prone to errors. Furthermore, understanding the data usually requires the knowledge of qualified specialists. These restrictions may make it more difficult to detect diseases promptly and increase the chance that the disease will spread across the flock, which could have dangerous results. Methods: In this paper, we use the PomeNetV2 Convolutional Neural Network (CNN) architecture and also examine 4 diseases of pomegranate fruit with names: Alternaria, Anthracnose, Bacterial Blight, Cercospora and Healthy. We used a proposed dataset of 5099 pomegranate fruit disease images. We compared the results of the proposed pomegranate fruit disease classification method with those of existing works. Result: Based on our experimental findings, the suggested framework outperforms all other state-of-the-art models with an accuracy of 99.02% for 75 epochs in identifying healthy and diseased pomegranate fruit.