Rice Blight Disease Intensity Classification with CNN-SVM Fusion

The production of rice is an important factor in both the food security and the economic stability of the world. Rice crops, on the other hand, are subject to a wide variety of illnesses, including blight diseases, which, if not handled properly, can drastically lower output. The proposed study provides a cutting-edge classification system for rice blight disease in this body of work. The system attempts to differentiate between five separate severity categories, namely Healthy, Mild, Moderate, Severe, and Critical. Convolutional Neural Networks (CNN) are utilized in the cutting-edge method for feature extraction, and Support Vector Machines (SVM) are utilized for final classification. This approach allows us to capitalize on the benefits offered by both methods. For training and evaluating the model, a dataset of 20,000 high-resolution photos that had been rigorously curated was compiled and annotated. The fusion model was successful in disease categorization, as evidenced by the fact that it attained an outstanding overall accuracy of 97.9%. In addition, complete performance metrics such as precision, recall, and F1-scores were computed for each disease severity level to demonstrate the consistency and dependability of the model. The confusion matrices provide insights into the model's behavior, and the graphical comparisons with other models demonstrate the superiority of the model. The accuracy of the model that has been proposed is exceptional, and in addition, it offers farmers and other agricultural specialists a useful tool that enables early disease identification and accurate assessment of the disease's intensity.