Collaborative Intelligence in AgriTech: Federated Learning CNN for Bean Leaf Disease Classification

Abstract

This study introduces a convolutional neural network (CNN) technique based on federated learning to classify bean leaf diseases. The research allays data privacy by allowing users to train local models on their datasets without sharing raw data. Our method combines local models from four customers to produce a high-performing global model that can categorize bean leaf diseases into five groups. According to the findings, the local models for each customer performed well in terms of precision, recall, F1 score, and accuracy. The performance measures’ macro, weighted, and micro averages showed that the aggregated global model performed equally well across all customers. The global model’s weighted average precision, recall, F1-score, and accuracy were 92.61%, 89.72%, and 92.87%, respectively. This research demonstrates how the federated learning-based CNN technique can effectively use various data types from multiple clients while maintaining data privacy. The accuracy with which this method identified bean leaf illnesses demonstrates the potential of federated learning in the agricultural field and provides a viable strategy for further study and real-world applications.