Improved Food Security of Bottle Gourd Leaf Disease by Federated Learning CNN: A Case Study

This study uses Convolutional Neural Networks (CNN) in a networked machine learning framework called federated learning to research classifying six-bottle gourd leaf illnesses. In this study, four more clients build a federated learning model, each hosting a fraction of the total data. The goal is to provide a thorough, expandable, and privacy-preserving system for locating and classifying Bottle Gourd leaf ailments. This will enable the control of diseases more effectively, ultimately leading to higher agricultural yields. The model performed well in analyzing client outcomes across several subsets of data, with high accuracy values ranging from 96% to 98% across various disease categories. Furthermore, each sickness class's accuracy, recall, and F1-score displayed high values, demonstrating the model's ability to identify and classify illnesses accurately. To study the overall results, the performance of the federated learning model was evaluated using consolidated data. It demonstrated high accuracy rates, ranging from 89.51 % to 91.39%, with similar recall, F1-score, and support statistics. Its consistently high accuracy showed the model's ability to generalize across the whole data universe. The global model's macro, weighted, and micro average precisions are carefully calculated in this study to assess it. A strong performance and consistency in illness categorization are shown by the model's high accuracy values across all averages for each client. The results of this study provide a federated learning-based method for disease identification in Bottle Gourd leaves, which significantly contributes to intelligent agriculture. The paper also emphasizes how federated learning can be scaled up while maintaining privacy, indicating that it might be appropriate for use in more general agricultural situations.