PaddyShield: A CNN-Based Rice Leaf Diseases Detection and Management System

Sri Lankan paddy farmers face numerous challenges due to limited access to resources and knowledge of pests and diseases effectively. This often leads to wasting money on pesticides without addressing the problem, affecting crop yield. Existing systems excel at swift disease detection but do not acknowledge farmers in preventing and mitigating diseases. Farmers expect information on the diseases and suitable mechanisms to get rid of the diseases. Further, we conducted a pre-survey with paddy farmers in the Anuradhapura district. We understood the importance of acknowledging the pests and diseases and the methods to mitigate them. Therefore, we developed a “PaddyShield” system to identify diseases and provide comprehensive disease details on prevention mechanisms for the identified diseases through rice leaf. The “PaddyShield” system was implemented after employing the sophisticated Convolutional Neural Network (CNN) as the linchpin for precise disease detection. The system’s strength lies in its adept handling of a diverse dataset encompassing various paddy leaf diseases, ensuring accuracy ($\mathbf{9 5 \%}$) in early disease identification. Through rigorous testing, the system indisputably highlights its superiority over existing solutions. Accessibility is prioritized through a user-friendly Flaskpowered interface. Overcoming hurdles like data preprocessing and model complexity with inventive solutions, the system achieves seamless backend-frontend integration. Optimal GPU resource utilization further accentuates its efficiency.