Unleashing the Potential of Federated Learning and CNNs in Potato Leaf Disease Detection and Severity Evaluation

This study is the first to employ convolutional neural networks (CNN) and federated learning to identify and categorize six serious potato leaf diseases. The novel methodology used in this work protects data privacy while improving illness detection and classification precision. Four local client datasets were analyzed thoroughly, resulting in reliable measures. Specifically, precision varied from 89.77% to 94.92%, recall varied from 88.44% to 95.15%, the F1-score ranged from 91.77% to 93.75%, and accuracy stayed consistently high, peaking at 0.98. These results, which also applied to the globally aggregated data from these customers, further demonstrated the model's dependability and robustness across various datasets. Precision, recall, F1-score, and accuracy scores for each customer varied from 92.66% to 93.53%, demonstrating high consistency. Finally, analysis employing Macro, weighted, and micro averages, scoring over 92 % in all measures, validated the model's overall effectiveness. This demonstrates the model's skill in managing the contributions from all classes and maintaining class balance. In summary, this work highlights the tremendous potential of federated learning in conjunction with CNNs to provide a trustworthy, effective, and privacy-preserving solution for detecting potato leaf disease and grading its severity. In summary, the work effectively uses federated learning capacity to train a CNN to diagnose and categorize the severity of potato leaf disease. The study highlights the potential for technological breakthroughs to transform crop disease control and boost agricultural output. It significantly contributes to data privacy, advanced machine learning, and agriculture.