A New Age for Agricultural Diagnostics: Federated CNN for Papaya Leaf Diseases

The growing problem of papaya leaf diseases has prompted a pressing need for effective remedies. As a reaction, the study utilizes the emerging capabilities of convolutional neural networks (CNN) and federated learning to provide a distinct, decentralized, and all-encompassing diagnostic method. Using five different classes of Papaya leaf diseases, the technique examined and analyzed each local data repository independently for each of the five customers (cy_1 through cy_5). After extensive training and validation, several statistical discoveries were made. The precision values obtained from the local data result analysis were as follows: 80.32% for cy_1-fe_1 and a staggering 90.37% for cy_5-fe_5. The accuracy, recall, F1-score, support, and accuracy values further validated the robustness and flexibility of the federated CNN model. Extending an integrative perspective, the research used Federated Averaging methods (Macro, Micro, and Weighted Averages) to convert local knowledge into a comprehensive worldwide outlook. With figures like 83.10% in Macro, 83.76% in Weighted, and 83.64% in Micro averages, cy_5 demonstrated unmatched numerical dominance, although other clients maintained vital metrics like 80.46% (cy_1 Macro) and 78.93% (cy_4 Micro). Federated Averaging was essential in turning local data into global data at this crucial stage. Client-specific information was combined to create a harmonized, universally applicable model, such as the 80.70% accuracy and 83.39% recall of cy_1. The work provides a new diagnostic tool for papaya leaf diseases. It establishes a standard for using decentralized data in conjunction with CNNs and federated learning to address domain-specific problems.