Enabling Precision Agriculture: A Federated Learning CNN Approach to Diagnosing Jackfruit Leaf Diseases

This study examines the use of Convolutional Neural Networks (CNN) with Federated Learning to diagnose five different types of jackfruit leaf diseases. This work develops federated architecture aggregating model changes to provide a global model, maintaining correctness and privacy and data security. It does this by using local data from five different clients. Performance indicators, including precision, recall, F1 score, and accuracy, were meticulously scrutinized in the context of result analysis. Client r-1, for instance, displayed an accuracy of 0.98 and precision, recall, and F1 score values of 94.47%, 94.53%, and 94.49%, respectively. These parameters, which show strong effectiveness, go up to 98.01% precision, 98.03% recall, 98.02% F1 score, and 0.99 accuracy in client r-3. The study further details federated averaging to convert local data to global data. We investigate the macro, micro, and weighted averages. All customers' Macro Averages swing between 94.50% (r-1) and 98.02% (r-3). Weighted Averages, which range from 94.65% (r-1) to 98.04% (r-3), roughly match these numbers. Micro Averages are also in agreement, ranging from 94.65% (r-1) to 98.04% (r-3).In conclusion, this study highlights the robustness and generalizability of federated learning using CNNs to identify illnesses in jackfruit leaves while proving its viability. The model's efficiency is supported by its high values in the Macro, Micro, and Weighted Averages, making it a potential area for more research and practical implementation.