An efficient deep learning approach to detect citrus leaves disease

Abstract

Bangladesh is one of the leading exporters of citrus. The country has been exporting citrus fruits to more than 60 countries annually. The main risk that citrus disease poses to crop yield is through contact with infected fruit. Identifying different dis eases of citrus leaves needs a huge time, work, and expertise. As a result, a new citrus disease detection technology must be developed. Infected crops need to be harvested as soon as possible before they rot. We have developed a useful technique in this study to use deep learning models to detect illness in citrus leaves. Using a unique ensemble approach, we are now able to train the model with different numbers of classes, excluding the best illnesses, and then worked together on the forecast. Each plant’s state is determined by taking a snapshot of its leaves and analyzing them. Data collection, pre-processing, segmentation, extraction, and classification are used to detect leaf disease. In this study, plant diseases were identified using photos of their leaves and segmentation and feature extraction algorithms. Our method can predict illnesses with an accuracy of 95% by combining many classifications, which represents a significant opportunity to save production losses.