Different Leaf Disease Detection by Deep Learning Approach

Abstract

The use of deep learning techniques to identify common leaf diseases in different crop species is the main focus of this study, which makes use of a dataset of 10,000 unprocessed mobile phone photos. The study encompasses a selection of plant species, namely Grape, Lychee, Peach, Pepper, Strawberry, and Potato, each afflicted with prevalent illnesses such as Grape esca, Lychee twig blight, Peach bacterial spot, Pepper bell bacterial spot, Potato late blight, and Strawberry leaf scorch. The study employed three distinct CNN models. Model 2 had suboptimal performance, while Models 1 and 3 demonstrated high levels of accuracy. The accuracy rate of Model 3, specifically, was notable at 94%. The suggested models were evaluated using a comprehensive dataset consisting of both healthy and injured leaves. The experimental design employed in this study was the capturing of sound and its subsequent impact on the leaves of specifically chosen trees, with the aim of assessing its potential for practical implementation in real-world scenarios. The third iteration of deep learning models demonstrates promise in achieving precise and efficient identification of leaf diseases across diverse agricultural settings. These findings facilitate the timely detection and management of plant diseases, hence enhancing crop yield and promoting agricultural sustainability.