A Paradigm Shift in Pomegranate Leaf Disease Detection with Federated Learning and CNN

Abstract

Pomegranate leaf diseases offer severe obstacles to agricultural output, calling for quick and efficient means of identification. The convolutional neural networks (CNNs) and federated learning techniques used in this article to identify pomegranate leaf diseases are innovative. The primary goal of this project is to implement a high-performing distributed machine-learning approach that protects data privacy. Each of the four customers used to simulate the issue dealt with six distinct kinds of pomegranate leaf diseases. Several performance criteria, including precision, recall, F1-Score, and accuracy, assessed the approach. Using the study, to were able to identify customers who had improved their performance over time, leading to remarkable precision, recall, and F1-Scores of 96.19%, 96.86%, and 96.53%, respectively, with Client C having an accuracy rate of 98%. The research examines global performance, combining local knowledge with a worldwide model. On Client C, this federated strategy produced significant precision, recall, F1-Score, and accuracy scores of 95.03%, 94.91%, 94.96%, and 98%. The values were notable for remaining high across all customers, demonstrating the model's robustness and durability.Additionally, macro, weighted, and micro averages were used to examine the model's performance, thoroughly understanding how well it performed across all classes. With a macro average of 94.96%, a weighted average of 95.07%, and a micro average of 95.06%, Client C showed the highest level of performance. The results support the potential of this strategy to provide farmers and other agricultural stakeholders with a dependable, highly effective, and privacy-preserving solution, thereby promoting increased production and sustainable farming practices.

Topics & Concepts