Smart Agriculture: Innovating Soybean Leaf Disease Detection with Federated Learning CNNs

This paper presents a novel study on the classification of soybean leaf diseases using federated learning with convolutional neural networks (CNNs). Based on five different classes of soybean leaf diseases, this study relies on the data from five clients to represent various environmental conditions, forms, and variations in diseases equally. One central aspect of this research is the federated learning approach, through which distributed data can be combined while protecting people's privacy and data. The basic premise of our analysis is the federated averaging process for macro, micro, and weighted averages that transforms local data understanding into a single global model. The performance of this model was quantitatively evaluated by precision, recall, F1-score support, and accuracy. The results also demonstrate the effectiveness of federated learning for more complicated, multiple-class classification tasks. For instance, the macro averages over clients ip1, ip2, ip3, and so on are 93.76% and 93.49%, respectively, from which it is clear that, combined with other factors, human labour accounts for a high proportion of costs at this particular plant concerning these three items: pills in powder form or oil injections directions finished products to. Also, weighted averages appear as 93.75 %, 93.48 %, 94.48 %, and 92.36 % and the micro standard is still set at its old levels. For example, qualifications were assessed from Teachers Certificate-University graduates by Institute type (A, B or C) according to classes. This study offers a systematic and quantitatively rigorous analysis of soybean leaf disease classification using federated learning and CNNs.