Aloe Vera Leaf Diseases Pathology: Harnessing Federated Learning CNNs for Enhanced Detection

This research paper proposes a novel technique that utilizes the federated learning framework and convolutional neural networks (CNN) to identify aloe vera leaf diseases. The study targets five main species of aloe vera leaf diseases using data from clients in about five countries. The ultimate goal is to use federated learning techniques for aggregation while achieving data privacy and minimizing the need for centralized data when the model is rigorously tested using precision, recall, and F1-score. Support and accuracy metrics to calculate its performance The clients effectively detected disease to different degrees, which indicates a geographically heterogeneous manifestation of the said condition. The performance metrics across the clients were as follows: Precision, recall F1-score and accuracy of 81.54%, 81.59 %, and 80% were demonstrated by the client. The precision, recall, F1-score and accuracy of the client rj_2 were 79.03% and 79.19%, respectively, as well as a total of precisely The precision, recall f-score, and accuracy scores for the client rj_3 were 87.05%, followed by 87.10%, followed up with 87.07%, and finally ending at 95%. Secondly, the investigation focused on evaluating federated averaging for model aggregation using local client data and comparing it with global aggregate data. The analysis revealed a progressive enhancement in model accuracy and reliability, as evidenced by the average performance metrics: Macro averages were 81.56% (rj_1) and 90.02% (rj_5). The weighted average scores ranged from 81.63% (rj_1) to 90.01% (rj_5). The micro averages went from 81.59% (rj_1) to 90.02% (rj_5). These findings support the model's ability to respond to region-specific diseases and improve its performance by employing federated learning. The approach is instrumental in agricultural settings where the data are fragmented and heterogeneous.