1Department of Electronics and Communication Engineering, Ballari Institute of Technology and Management, Ballari, Affiliated to Visvesvaraya Technological University, Belagavi-590 018, Karnataka, India
2Department of Electronics and Communication Engineering, Rao Bahadur Y. Mahabaleshwarappa Engineering College, Ballari, Affiliated to Visvesvaraya Technological University, Belagavi590 018, Karnataka, India
*Corresponding Author: Ambrayya, Department of Electronics and Communication Engineering, Ballari Institute of Technology and Management Ballari, Affiliated to Visvesvaraya Technological University, Belagavi-590 018, Karnataka, IndiaEmail: ambrayyabs@gmail.com
Paddy crops face severe threats from multiple pathogens that attack leaves, stems and panicles, causing massive yield drops and compromising food security. Identifying these diseases quickly and accurately is crucial for effective treatment.
This research introduces an Optimised Region-Based Fully Convolutional Network (RFCN) built to detect and localize several paddy diseases at the same time. By leveraging a region proposal network (RPN) alongside advanced feature extraction and multi-class classification, the system pinpoints infections at the organ level, even when different diseases coexist on a single plant. Training and testing relied on a carefully annotated image dataset of paddy specimens.
The optimized RFCN recorded a mean Average Precision (mAP) of 88.7%. When compared against standard architectures like Faster R-CNN and SSD, the proposed model showed superior precision, recall and localization. It processes images at 11 FPS, positioning it as a viable tool for real-time agricultural diagnostics.
Computer vision, Multi-disease detection, Object localization, Precision agriculture, RFCN
