Innovative Approaches in Agri-tech: Federated Learning Models for Disease Detection in Papaya Leaves

Diseases particularly affect the leaves of papaya plants significantly affect yield as they reduce both quantity and quality. It is, therefore, important for agriculture to seek ways of solving the issues of fast and accurate methods of disease diagnosis. To achieve the goals of this work, which is the classification of the four main diseases affecting the papaya tree and its leaves, this paper proposes a new scheme that is based on federated learning, CNN, and the FedAvg algorithm. In contrast to most known approaches, by working with six different geographic sites to compute six individual models that contain locally relevant data, the presented federated learning scheme allows for fully efficient, scalable, and protective computation without the need for data unification, with sound indicators of 91% accuracy, which is also commonly known as precision, 90% recall, which is also acknowledged as sensitivity and 90 F-Measure. The overall response accuracy to the illness categorization was 92%, of which the F1-score indicated that this model was very effective, especially at 5% F1-score. In addition to minimizing major privacy issues, the amount of displayed data was reduced to a level in which the federated strategy helped decrease training bandwidth by as much as 76% for different models. Due to these features, the proposed approach is most suitable for actual agricultural environments where data privacy and network performance are paramount. Our study shows that federated learning can revolutionize agricultural disease management due to its capability of offering timely and precise interventions.