Rice Leaf Disease Detection and Diagnosis Using Convolution Neural Network

Agriculture is the art and science of soil cultivation, crop production, and livestock management. The majority of the world’s food is produced by agriculture, where paddy fields are most often cultivated crops and its diseases are the main problem of its loss. Rice is an example of a global crop, and India is ranked second in rice production. A technique that detects the leaf disease will greatly help farmers and agriculturists. The agriculture industry must overcome this loss due to the affected leaf diseases that bring down the production. As per studies, it is found that various types of diseases affect the leaf, such as infectious and non-infectious. This paper proposes a method to detect and diagnose rice leaf diseases with appropriate remedies. This research has considered five diseases – Leaf Blast, Hispa, Tungro, Leaf Blight and Brown Spot. Input images are segmented manually by removing their background if too many noises are seen. The model predicts the type of disease that affects the rice plant from an uploaded image or an image freshly clicked using a smartphone. Further, some preventive measures to overcome those diseases are suggested. Our model gave a sensitivity of 94.68% and a specificity of 66.66% resulting in an overall accuracy of 93%.