Federated Learning Based Crop Disease Detection in Precision Agriculture

Crop diseases significantly impact agricultural productivity, necessitating early and accurate detection to minimize yield loss and enhance sustainability. This study presents a federated learning-based framework for detecting diseases in staple crops such as rice, wheat, and potatoes. By integrating deep learning models like MobileNetV2 and InceptionV3 within a federated learning setup, the proposed framework addresses limitations of traditional methods while preserving data privacy. A dataset of 4,500 images, spanning three crops each with two diseases and one healthy category, was used to improve model robustness and accuracy. Federated learning with MobileNetV2 achieved an accuracy rate of 97.00% by effectively identifying diseases of potato, wheat and rice. The framework’s advanced feature extraction and transfer learning capabilities enable efficient and real-time disease detection. The experimental results highlight the potential of federated learning to support sustainable agriculture and global food security by reducing crop losses and adapting to diverse local conditions while preserving data privacy.