An Ensemble Learning Approach for Early Detection and Classification of Plant Diseases

Plant diseases are a major threat to agriculture, particularly impacting the yield and quality of tomato crops. Early detection and accurate classification of these diseases are essential for effective management and mitigation. Traditional methods of disease detection are often labor-intensive and time-consuming. Although individual convolutional neural networks (CNNs) have shown promise in automated plant disease detection, their accuracy and robustness can be limited when used in isolation. This study proposes an ensemble learning approach that combines three state-of-the-art CNN architectures: AlexNet, ResNet50, and VGG16. A comprehensive dataset of tomato leaf images, categorized into bacterial, viral, fungal diseases, and healthy leaves, was used. Images were preprocessed and augmented to improve model generalization. Each model was trained separately, and their outputs were integrated using a weighted averaging mechanism to form the ensemble model. The weights for each model were optimized based on validation performance. The ensemble model significantly improved classification accuracy compared to individual models. The combined approach achieved an overall accuracy of 97.5%, with precision, recall, and F1-score exceeding 95% for all disease categories. Specifically, the accuracy for detecting bacterial diseases was 96.8%, viral diseases 97.2%, and fungal diseases 97.9%. The ensemble method demonstrated superior robustness and reliability in classifying diverse disease symptoms.