Deep Ensemble Learning for Rice Plant Disease Identification with Proposed Control Measures

Rice is an indispensable staple crop crucial for ensuring global food security, especially in Asia. Detecting rice plant diseases early is crucial to prevent widespread damage and ensure high yield and quality. In recent years, deep learning and computer vision techniques have greatly improved the automatic identification of rice plant diseases. In this study, an ensemble technique is proposed that combines ResNet-9, VGG-19, and DenseNet-121 models for detecting eight different types of rice plant diseases. The dataset comprises images representing eight distinct rice plant diseases, along with images of healthy plants. The individual models—ResNet-9, VGG-19, and DenseNet-121—achieved accuracies of 85%, 97%, and 97%, respectively. However, after applying the ensemble method, the overall accuracy improved to 97.56%. These outcomes underscore the effectiveness of the proposed ensemble approach in accurately identifying rice plant diseases, contributing to sustainable agriculture through timely and precise interventions. Finally, a user interface is developed that would suggest disease name and necessary control measures for each type of disease.