Elevating Aloe Vera Leaf Disease Detection through Federated Learning CNNs

This study explores the novel use of federated learning in conjunction with CNN for illness classification in Aloe vera leaves. Five separate client-contributed classes of Aloe Vera leaf diseases form the basis of the dataset used in the research. The technique is centered on using federated learning to create a decentralized data analysis system that protects privacy. Each client's local datasets are used to create a model that is optimized worldwide. The examination of the results included a thorough evaluation of the model's performance across all five customers (nh_1 to nh_5) using a variety of statistical metrics. To provide a complete picture of the model's performance, compute the macro, micro, and weighted averages. Results showed that nh_1, nh_2, nh_3, nh_4, and nh_5 all had macro standards of 88.61%, 84.07%, 80.08%, 90.02%, and 94.16%, respectively, indicating that the model performed evenly across all classes. Using weighted averages, which take into account the class imbalance, further analysis revealed that nh_5 was in the lead at 94.25%, followed by nh_4 at 90.04%, nh_1 at 88.97%, nh_2 at 84.45%, and nh_3 at 80.13%. Micro averages, which take into account all classes' contributions, con firmed similar findings: NH 5 achieved 94.25%, NH 4 gained 90.04%, NH 1 reached 88.94%, NH 2 achieved 84.49%, and NH 3 achieved 80.15%. In particular, for agricultural disease identification, the study's major contribution is the way it shows how federated learning may be used to merge local data into a strong global model.