Enhancing Sugarcane Leaf Disease Diagnosis with USEM: A Hybrid Approach Using U-Net and Stacked Ensemble Learning

Abstract

The precise identification of sugarcane leaf diseases is a key factor in minimizing agricultural losses. The work aims to propose a robust approach that combines U-Net segmentation with stacking ensemble learning. The U-Net architecture is utilized to segment diseased regions, and features extracted from these segments serve as input for training base learners, including Decision Trees, Support Vector Machines (SVM), and Convolution Neural Networks (CNN). By aggregating predictions from these base learners using a meta-learner, we achieve improved classification performance. Evaluation metrics such as accuracy, precision, recall, F1-score, and Intersection over Union (IoU) are employed, with cross-validation ensuring model robustness. Our methodology outperforms existing models, providing an effective strategy for early disease diagnosis in sugarcane and promoting sustainable agricultural practices.