Severity-Level Assessment of Groundnut Leaf Diseases: A Federated Learning and CNN Approach

Early diagnosis and categorization of plant diseases, especially those that affect groundnut leaves, are crucial in agriculture. This study examines cutting-edge-based Federated Learning across five different data clients, each representing a different plant leaf disease severity level (1-20%, 21-40%, 41-60%, 61-80%, and 81-100%). The local performance of the model was assessed for each of the five customers, Cl_j, Cl_f, Cl_h, Cl_l, and Cl_g. The precision, recall, F1-Score, and accuracy of Cl_j (indicating 1-20% illness severity) were 80.08%, 85.20%, 82.56%, and 0.96, respectively accuracy for Cl_g (81-100% severity) were 96.33%, 96.25%, 96.29%, and 0.98, improving performance with increasing illness severity. The Global Results Analysis, which represents the overall effectiveness of the federated learning model, revealed high precision, recall, and F1-scores consistently over 87%, as well as accuracy levels above 96% for all customers. The highest severity level, Client G, in particular, produced almost comparable precision, recall, and F1-score values, each around 94.8%, with a high accuracy of 98%. Macro, weighted, and micro averages all increased steadily from Client J (macro: 87.69%, weighted: 88.86%, micro: 88.83%) to Client G (macro: 94.84%, weighted: 94.89%, micro: 94.89%) in the Comparative Analysis of Averages. The robustness and dependability of the federated learning model in recognizing and categorizing the severity levels of groundnut leaf diseases are shown by these averages across all measures. The federated learning models’ consistently excellent performance across various severity levels highlights their promise as a flexible tool for addressing agricultural data’s varied and dispersed nature.