Bridging the Strawberry Cultivation Gap: Federated Learning CNN for Disease Detection

Abstract

Leaf infections are a persistent problem in strawberry agriculture, an agricultural sector with significant economic importance. Five primary kinds of leaf diseases significantly reduce productivity and quality. To identify and classify these disorders, this study presents a novel method that combines the image classification skills of convolutional neural networks (CNNs) with the decentralized nature of federated learning. The research can capture various illness presentations by working with five customers, each representing a distinct regional and cultural context. Several performance metrics are shown for each client in the result analysis: Py_1 performs very well with an accuracy of 0.96, closely followed by Py_5 at 0.98, while Py_2,Py_3, and Py_4 record accuracies of 0.94, 0.93, and 0.93 and so on. Federated averaging is key in moving from localized models to a global viewpoint. Py_5 was the clear winner when the averaging measures were examined. Its macro, weighted, and micro averages were 94.72%, 95.14%, and 95.14%, respectively. Py_3, in sharp contrast, falls significantly behind the other two averages, with values of 83.02%, 83.45%, and 83.52%. This work provides a strong, decentralized model for classifying strawberry leaf diseases and highlights the effectiveness of federated learning in integrating various data sources. A bright future for practical applications in agricultural diagnostics is predicted by transforming local data viewpoints into an all-encompassing global model, as shown by the constant performance metrics after federated averaging.

Topics & Concepts