Deep Learning Insights into Banana Sigatoka Disease: ResNext50 for Seriousness Classification

The global cultivation of bananas is significantly threatened by the presence of Banana leaf sigatoka disease, which is predominantly caused by fungal diseases known as Mycosphaerella fijiensis and Mycosphaerella musicola. The precise evaluation of illness Seriousness is of utmost importance to ensure the efficacy of disease management strategies. The present study utilizes the ResNext50 deep learning (DL) model to construct an automated classification framework aimed at distinguishing the Seriousness levels of Sigatoka illness in banana leaves. Our study is based on a rich dataset consisting of 10,000 photos of banana leaves. These images have been annotated with five Seriousness levels, providing a detailed and extensive resource for our research. By employing rigorous data preprocessing techniques and doing thorough fine-tuning of the ResNext50 model, our system attains an overall accuracy of 95.53%, as evidenced by the achieved accuracy percentage. Evaluation indicators such as accuracy, recall, and F1-Score offer a detailed comprehension of the performance of the model. The confusion matrix provides valuable insights into the predictive performance of the model for each class. The model's competitive advantage is highlighted by a comprehensive graphical comparison of its state-of-the-art features. The present study not only focuses on the immediate issue of Sigatoka disease but also establishes a foundation for the implementation of automated disease evaluation in precision agriculture, thereby making a significant contribution to the sustainable management of crops.