Predicting Cotton Leaf Disease Based on Image Classification Using Deep Learning

Abstract

Cotton is one of Bangladesh's most important agricultural products, but it faces a number of challenges or constraints in the leaf. Most of the time, these constraints are identified as diseases and pests that are difficult to detect with the naked eye. The goal of this study was to create a model to improve cotton leaf disease detection and prediction using CNN, based a deep learning technique. This study used a raw dataset containing bacterial blight, curl virus, fusarium wilt, verticillium wilt, and healthy leaves to accomplish this. The dataset split into 80:20 which boosted the generalization of the CNN model. For this research, the dataset has nearly 2535 images where 80% were accessed for training purposes. This developed model is implemented using python version 3.9.13, and the model is equipped with the deep learning package called Keras, TensorFlow, and Jupyter which are used as the developmental environment. This model achieved an accuracy of 96.88% for identifying classes of leaf disease in cotton plants. This paper aided the agricultural sector in transitioning away from traditional or manual disease and pest detection methods in order to achieve breakthrough results. Large farms will greatly benefit from this automated process for reducing monitoring work. Keywords: Deep learning, Convolutional Neural Network, Cotton leaf diseases.