Empowering Vineyards: Decentralized Grape Leaf Disease Recognition with Federated Learning CNN

To accurately identify and categorize grape leaf diseases, to offer a Federated Learning Convolutional Neural Network (FL- CNN) model in this research study. Regarding productivity and quality, grapevine diseases substantially influence the grape business. Rapid and accurate disease diagnosis is essential to implement efficient illness management strategies. Traditional machine learning techniques sometimes need a centralized dataset, which raises questions about data security and privacy. The work uses federated learning to address this problem by allowing several clients to train the CNN model while maintaining data privacy jointly. It is assessed on a diverse dataset of 6,798 images of grape leaves, representing five classes representing four grape leaf diseases and one healthy class. Four local clients participate in the federated learning process, each enhancing the performance of the overall model. The results demonstrate the FL-CNN model's potency in identifying and categorising grape leaf diseases. To illustrate the global model's overall performance balance and efficacy at both the client and individual sample levels by examining the Macro, Weighted, and Micro averages ranging from 91.46% to 92.88%. In conclusion, when paired with federated learning, the proposed FL-CNN model provides a reliable and privacy-preserving approach for precise grape leaf disease detection and classification. This strategy has the potential to significantly improve disease control tactics used in the grape sector, eventually leading to better yield and quality.