Agriculture Innovations: Broccoli Leaf Diseases Classification through Federated Learning CNN's

The study concentrates on applying federated learning with convolutional neural networks (CNN) to enhance the accuracy of diagnosing and classifying broccoli leaf diseases in agriculture. This paper provides details of a distributed machine-learning model applied to four agricultural customers to collaboratively analyze plant pathology, focusing on four severity levels of leaf disease. The experiment was based on federated learning, which integrated the local data prophecies into a powerful global model. CNN was the core analysis of the model for classifying the diseases by images. The research found that the federated CNN model achieved very high accuracy, ranging from 92.76% to 97.56% for the customers and recall from 92.83% to 97.55%. The F1 scores, which can control accuracy and recall, varied from 92.78% to 97.55% and were always high. The support values, which denoted how many actual instances of each class were, revealed a significant data representation ranging from 391 to 721.75 instances. Consistency showed exceptional model accuracy for all customers, never dropping below 96%. The validity of a model integrating macro, micro, and weighted averages was verified by thorough research that covered local and global data patterns. The macro averages fluctuated between 92.79% and 97.55%, which shows the consistency of the model in every category. The weighted averages, considering class imbalances, recorded high values varying from 92.85% to 97.54%, indicating the model's capability to deal with varied class distributions. The micro averages were very close to the weighted averages, with the range being 92.84% to 97.54%, and thus, the model provided a consistent performance in all categories.