Agricultural Insights Through Federated Learning CNN: A Case Study on Jackfruit Leaf Disease

Abstract

Particularly in areas where jackfruit is a significant crop, the rising frequency of jackfruit leaf diseases severely threatens agricultural productivity and sustainability. Convolutional neural networks (CNNs) are used in this study's novel federated learning model to examine the severity of jackfruit leaf diseases across six diverse clients with various degrees of data richness. This federated architecture preserves data privacy and lowers data transfer costs by allowing localised learning on clients and combining the local models into a global one. Four severity levels-minimal (1-25%), moderate (26-50%), severe (51-75%), and critical (76-100%)-are used to assess the model's performance systematically. The Analysis of Results Astonishingly accurate metrics are revealed by CNN using federated averaging; the Macro, Weighted, and Micro Averages varied from 88.65% in the client with the lowest performance (Z_1) to an unmatched 96.18% in the client with the highest performance (Z_6). The accuracy values of the model's statistical fulcrum typically remain between 94% and 98%, demonstrating strong performance. The federated averaging-derived aggregated global model highlights the model's unwavering stability and scalability in addition to supporting correctness. Therefore, this groundbreaking study is a cornerstone in developing precision agriculture and presents a possible answer to the growing problem of jackfruit leaf diseases.

Topics & Concepts