Deep Learning-Based Automated Detection of Cotton Plant Diseases: A Comparative Analysis of CNN Architectures

Abstract

Deep learning methods for crop disease detection and identification of pests on leaves can significantly improve agricultural productivity and greater reliance than manual inspection on a specific road map for growing cotton. In this study, we have begin by reviewing some existing methods using Convolutional Neural Network (CNN) architectures, such as AlexNet, VGG, ResNet and DenseNet, to classify the cotton plant diseases. A comparative analysis indicating accuracy, precision, recall, and F1-score demonstrates that ResNet-152 and DenseNet-264 achieve better performance than other models in terms of maximum classification accuracy. The results of the classification show that deep learning models are accurately differentiate healthy and diseased cotton leaves, and diseased and healthy cotton plants, paving the way for early-stage diseases diagnosis.