Uncovering New Possibilities for Rice Agriculture: A CNN-SVM Model for Accurate Identification of Severity in Bacterial Brown Spot Rice Leaf's Disease
Shiva Mehta, Vinay Kukreja, Satvik Vats
Abstract
This project aims to create a computer vision-based model for identifying and categorizing bacterial brown spot disease in rice leaves. If not recognized and treated right once, bacterial brown spots, a common condition of rice leaves, may drastically diminish crop production. However, many studies have yet to use computer vision techniques to detect and categorize this illness. The work fills this gap by suggesting a CNN-SVM-based model trained on 6432 photos of excellent and diseased rice leaves gathered from nurseries in Punjabi villages. The presented model achieved an accuracy of 99.826% and 95.247% in binary and multi-classification, respectively, and showed excellent accuracy in differentiating between four distinct degrees of bacterial brown spot disease severity. The model performed very well at detecting diseases with moderate severity.Further proving its superiority in identifying and categorizing this illness, the suggested model served better than other previously trained models in diagnosing bacterial brown spots in rice leaf disease. The creation of this model has significant ramifications for rice leaf bacterial brown spot disease management and prevention. Farmers can quickly take action to control the illness and avoid considerable crop losses by having a reliable and precise approach to identifying and diagnosing this disease. This research also lays the groundwork for creating hybrid disease detection and classification models that may be used to recognize other illnesses in rice leaves. Overall, this study advances the application of computer vision techniques for disease detection and classification in agriculture, which benefits sustainable rice production and food security.