Real Time Analysis of Aloe Vera Disease Classification using Deep Learning

Objectives: To identify the plant diseases of Aloe vera (A. vera) in real time for early cure of diseases, an automatic system is designed to improve the plant yield. The objectives of this work are to develop a classification system of healthy and diseased A. vera leaf classes and to develop a user interface to support the real time analysis of the affected leaf diseases. Methods: To address the need of disease classification model, A. vera dataset from different sources are considered, both from open source and real time. Different deep learning models- Convolutional Neural Network (CNN), MobileNet(MNet), EfficientNet(ENet) are considered for classification of A. vera leaf diseases. A user friendly interface is hosted on a web platform that allows users to upload the clicked photos with their cameras of A. vera leaves for real time analysis and identification of diseases. Findings: The classification accuracy using a dataset of 9000 images for training and testing the CNN model leads to 87% accuracy, 86% precision, 91% recall, and a 0.89 F1-score. The accuracy of classification with MNet and ENet are 99% for the database considered. Novelty: In this work, deep learning algorithms are considered for A. vera leaf disease classification. A real time web based user interface provides A. vera growers a quick solution for the diseases found in farms by uploading the images taken using mobile phone camera and to receive remedial suggestions. Keywords: CNN model, Deep learning, EfficientNet, Leaf disease classification, Medicinal plants, MobileNet