Sooty Mold Disease Detection on Cotton Leaves Using Deep Learning

Abstract

Cotton growing in Pakistan, one of the key crops in both the agriculture and textile sectors, is facing serious problems related to diseases, such as Sooty Mold, which causes significant reductions in yield and quality. Traditional disease detection methods are recognized as inadequate; therefore, this research exploits Convolutional Neural Networks (CNNs), especially VGG-16 and Xception models, to automate and improve the accuracy of detecting Sooty Mold’s disease in cotton leaves. This study utilized a dataset consisting of 695 training images and 251 validation images to demonstrate the ability of the models to distinguish between healthy and Sooty Mold-infected leaves with impressive accuracy. The VGG-16 model achieved a training accuracy of 99.43% and validation accuracy of 99.40%. The Xception model achieved a training accuracy of 99.6% and a validation accuracy of 99.55%, which were better than those of the VGG-16 model. In addition to theoretical model development, this research furthers its implementation into a practical solution through the development of an Android application. The development of this application, aimed at real-time detection of diseases, is a significant step forward in the process of merging technology and agriculture. Cotton crop growers are presented with a tool for the immediate action and management of the Sooty Mold. The implementation of deep learning models in a user-friendly mobile platform introduces a new method for precision agriculture and, with it, the possibility of transforming crop disease management practices. The success of these models, along with mobile technology, has set a new standard for agricultural innovation, thus contributing significantly to agroecological farming practices and the preservation of valuable cotton crops, not only in Pakistan but also in other countries.