Optimized Rice Disease Identification Using CNN and Random Forest Techniques

This paper proposes novel four categories of the combined model of Random Forest Classifier and Convolutional Neural Network for the detection and classification of rice leaf diseases. Even with developments in agricultural technology, certain illnesses seriously reduce the output of rice crops, thus more efficient diagnosis techniques are needed. The suggested approach seeks to solve this by using Random Forest for robust classification based on high-level feature extraction from pictures using CNN's capabilities for such purpose. Our approach uses many layers of convolution and pooling meant to transform pictures into a complete feature set, so the CNN component contains A Random Forest classifier, set with 15 trees and a depth of 8, then makes use of this feature set to precisely identify many forms of rice leaf diseases. Using a dataset of 3,245 photos including several forms of five main rice leaf diseases, the model was trained and evaluated. With an amazing total accuracy of 95.66%, the hybrid CNN-Random Forest model outperformed single CNN architectures like DenseNet, ResNet, and VGG16. These results not only highlight the possibilities of hybrid models in agricultural uses but also imply a scalable method for enhancing crop disease diagnostics throughout different areas, thereby supporting global food security.