Machine Learning Data Fusion for Plant Disease Detection and Classification

Abstract

It is crucial to quickly identify plant diseases since they impede the development of affected plants. Despite the widespread use of Machine Learning (ML) models for this purpose, the recent advances in a subset of ML known as Deep Learning (DL) suggest that this field of study has much room for improvement in terms of detection and classification accuracy. To identify and categorize plant diseases, a wide variety of established and customized DL architectures are deployed with several visual analysis methods. In this study, we use deep learning techniques to create a model for a convolutional neural network that can identify and diagnose plant diseases using very basic photos of healthy and sick plant leaves. The models were trained using an open library of 20639 photos that included both healthy and diseased plants across 15 different classifications. Some model architectures were trained, with the highest performance obtaining a success rate of 97.70% in detecting the correct [plant, illness] pair (or healthy plant). Due to its impressive success rate, the model is a valuable advising or early warning tool, and its technique might be developed to help an integrated plant disease diagnosis system function in actual production settings.