Disruptive Farming Intelligence: Federated Learning CNN for Broccoli Leaf Disease Classification

This study utilizes federated learning framework and Convolutional Neural Networks (CNN) to detect and categorize diseases affecting broccoli leaves. This study has used data from six different customers for training and testing the proposed model, ensuring a comprehensive and representative dataset. This study highlighted four levels of severity for broccoli leaf diseases. The experimental results indicate that the federated learning strategy is efficient for handling data from numerous clients while retaining confidentiality at the same time. Macro average scores were also consistent with the model's balanced performance across all classes; they increased from 91.58% for client XP_1 to a final of 96.2%. This demonstrates that the model's capability to classify severity levels in illnesses continues on an ascending trajectory. As for the weighted average, which considers class imbalances, the dynamics for different data distributions were illustrated by its successive increase from %. Likewise, the microaverage increased from xp_1 91.59% to xp_6 96. In conclusion of this study, federated learning with CNN may help diagnose illnesses in broccoli leaves. This shows the reliability and resilience of the model when applied to different datasets and severity levels. Federated learning holds much potential in precision agriculture since one client consistently does better than another with respect to accuracy and precision. The federated learning model plays a significant role in agricultural health management and monitoring since the way it manages and solves data from various sources is almost perfect. To benefit agronomists and farmers, this study has developed a powerful tool to make important decisions regarding crop integrity and the consequences involved in disease control.