Dissecting Parsley Leaf Disease Severity: A Federated Learning and CNN Perspective

This study uses a Convolutional Neural Network (CNN) model in conjunction with a Federated Learning (FL) framework to thoroughly investigate the identification of parsley leaf diseases. Four local datasets representing four customers and five different degrees of parsley leaf disease severity were used to evaluate the approach. Precision, recall, F1-Score, and accuracy measures were used to assess the model's performance. The model performed well for all customers in the local results and analysis, with an accuracy hovering around 0.98. The highest severity level often had the most excellent accuracy, recall, and F1-Score, indicating skill in spotting the worst parsley leaf infections. For example, Client 1 obtained precision, recall, and F1-Score values for severity levels 1-20% of 96.48%, 96.33%, and 96.41 %, respectively. The model showed good performance across customers, with accuracy ratings close to 0.98 in the area devoted to overall outcomes. The F1-Scores, which measure how healthy recall outperforms precision, were generally high among clients; Client 1 had an F1-Score of 95.73 %, for instance. Finally, the global averages section shows high-performance metrics for all averaging techniques. Precision, recall, F1-Score, and accuracy scores for the macro average were 95.73%, 93.98%, 94.83%, and 94.94%, respectively. Precision and accuracy were consistently close to 95% for both the weighted and micro averages, which showed comparable results. In conclusion, our study confirmed the value of using a CNN and federated learning strategy to identify illnesses in parsley leaves. High precision, recall, F1-Score, and accuracy are all shown in the findings, proving the model's dependability and robustness. The research confirms that our federated learning approach has the potential to be a helpful tool for farmers in controlling parsley leaf diseases, hence improving agricultural practices and enhancing food security.