Competent Diagnosis in Agriculture: Federated Learning CNNs for Avocado Leaf Diseases

This research paper explores the novel application of federated learning with Convolutional Neural Networks (CNN) for avocado leaf disease classification across the four severity categories. Based on the details from four unique customers, the analysis shows how federated learning can capitalize in agricultural environments around local data variations and improve the global model's accuracy. Our approach is the only one to tackle the classification issues with four disease severity classes. However, the fusion of federated learning with CNNs secures data privacy by processing locality at each client and promotes model generalisability over the various datasets. In the results, for each client (kg_1 to kg_4), there are signs of a good level of precision, recall, F1-score support, and accuracy in the disease classification. In particular, the client kg_1 determined a macro average of 96.73%, a weighted average of 96.80%, and a micro average of 96.80% with an accuracy equal to. Client kg_3 was optimized with a macro-average efficiency score of 98.95%, a weighted average for performance scoring set to 0.7, and a micro-average. The peak performance was finally reached for client kg4 with a macro average of 99.15%, a weighted average of 98.73%, and a micro average age of ninety-six while maintaining accuracy. These results show the high performance of the federated learning CNN model in diagnosing avocado leaf diseases with great precision and present many opportunities to use this methodology for other agricultural applications.