This paper presents a cost-efficient, AI-driven framework for real-time detection and alerting of rice crop diseases using YOLOv11. The proposed system addresses the limitations of existing solutions by leveraging existing low-cost field cameras or drone/UAV-captured images with geotagged metadata, eliminating the need for new camera infrastructure and reducing costs significantly. The YOLOv11 model is employed for multi-class detection of Bacterial Blight, Rice Blast, and Brown Spot, achieving a Box Precision (P) of 0.649, Box Recall (R) of 0.569, mAP50 of 0.626. The system integrates with a localized Interactive Voice Response (IVR) system for voice alerts and includes a field health scoring mechanism to provide actionable insights to farmers. The web-based dashboa rd health scores, upcoming risks like disease and alert logs, enabling farmers to monitor crop health remotely. Simulated case studies demonstrate the system's effectiveness in generating relevant alerts and tracking health scores over time. Future work will focus on improving model recall and mAP50-95 through further training with augmented datasets and expanding detection capabilities to include more diseases. This framework represents a significant advancement in AI-driven agricultural solutions for small-scale farmers, offering a powerful tool to enhance rice crop management and improve food security.
Download PDF: https://binra.eu.org/nmCiMH
