Cotton Disease Detection

Abstract

Cotton is India's most important crop, and maintaining the plants is crucial. Verticillium wilt, Alternaria spot, Cercosporin leaf spot, bacterial leaf blight, and red spot, are all diseases that harm cotton leaves. As a result, numerous methodologies have been introduced to aid farmers and increase crop productivity. Apart from those models, this paper outlines a fantastic method for detecting cotton plant diseases. A prior project-based effort is a web application. It describes a strategy that uses Partial Differential Equations (PDE)- based image decomposition, segmentation, feature extraction, feature selection, and classification to improve classification performance and propose a treatment plan. To partition the image into texture and object components, the total variation model is frequently utilized. The texture, color, and shape features are extracted using the codebook method and afterward combined into a feature set. The relief technique of selecting features is employed to keep only relevant attributes. Only a subset of the elements considered in classification is permitted to pass through the Multiclass classification Support Vector Machine (SVM) algorithm. Despite this, we developed a disease detection app using CNN, with a dataset of 2000 leaf images that incorporates the previously mentioned factors. The proposed technology will serve as a graphical interface for monitoring cotton leaf disease. We were able to identify the plants' diseases with 92.5 percent accuracy using the app.