An Advanced Deep Learning Approach for Precision Diagnosis of Cotton Leaf Diseases: A Multifaceted Agricultural Technology Solution

Abstract

During the past few decades, cotton leaf diseases have become a significant challenge for farmers, leading to substantial losses in harvests, productivity, and financial resources. Traditional observation methods are often time-consuming, costly, and prone to inaccuracies, exacerbating the plight of farmers in detecting and identifying diseases in their early stages. The consequences of late detection are dire, and both crops and farmers are under the brunt of prolonged infections. This study proposes a method to improve the detection of cotton leaf diseases by applying advanced deep transfer learning techniques. Using models such as ResNet101, Inception v2, and DenseNet121, and fine-tuning parameters utilizing the Nesterov accelerated gradient, the proposed system offers a powerful tool for farmers to swiftly and accurately diagnose leaf diseases. This system allows users to simply upload an image of a cotton leaf. After sophisticated image processing techniques, a Convolutional Neural Network (CNN) is deployed to detect the presence of cotton leaf diseases with high precision and efficiency. The experimental results demonstrated the effectiveness of transfer learning approaches, with the CNN achieving an impressive accuracy of 99%, while ResNet101, Inception v2, and DenseNet121 achieved 75.36%, 97.32%, and 97.16%, respectively. These findings underscore the potential of deep learning techniques to revolutionize disease detection in agricultural contexts, offering farmers a powerful tool to mitigate the impact of diseases on their crops.