Integrating CNN and SVM for Enhanced Classification of Maize and Sugarcane Leaf Diseases

Analyzing a crop accurately and managing its diseases precisely are required for increased agricultural output and assured food security. In image processing, CNN has emerged as a powerful and dominant technique. This research provides an approach to detect diseases in Kharif crops (sugarcane and maize) early on using hybrid deep learning architectures, such as Svm and Cnn. This research aims to blend valuable insights from hybrid crop identification and disease management techniques to foster sustainable farming practices. Maize is considered the third most important crop in India, contributing around 10% of total food grain production in the country whereas sugarcane contributes around 1.1% to India's GDP. Using visual data in the form of images, we have employed a hybrid strategy to categorize diseases identified in sugarcane and maize. Using a dataset of 14,000 photos, we have created 7 different classes, 4 of which are related to maize diseases and 3 to sugarcane diseases. Using CNN, high-level features have been extracted and classified using SVM. Training accuracy of the proposed model with the help of CNN used for feature extraction and SVM used for classification was obtained at 99.19%, validation accuracy was obtained at 82.64%, and test accuracy was achieved at 82.64%. High performance and outstanding discriminative capacity were shown through the confusion matrix and ROC analysis with an average AUC value of 0.92 across all classes. The hybrid technique shows prominent advantages over standalone CNN models with the help of possible applications in agricultural diagnostics and offers a feasible alternative for precise and effective finding of maize and sugarcane diseases.