Transforming Agriculture : Federated Learning CNNs for Wheat Disease Severity Assessment

Abstract

It is crucial to correctly identify and evaluate wheat illnesses to stop their growth and boost farming output. This research study intends to use CNN with collaborative learning to evaluate the seriousness of the wheat illness. Eight thousand nine hundred sixty-seven pictures of wheat plants with five distinct diseases—leaf rust, Fusarium head blight, wheat stripe mosaic virus, powdery mildew, and Septoria leaf blotch—were used in this research study. Federated learning, a learning method that enables numerous devices to train a machine learning model cooperatively while maintaining data privacy, was used to train the model. The model assessment findings demonstrated that the federated learning CNN model, with an average accuracy of 0.90, can correctly predict the intensity levels of various wheat illnesses. The model can correctly differentiate between the intensity levels of each disease, as evidenced by the high precision, F1-score, and accuracy for each of the five wheat diseases. Farmers and experts can take suitable action to stop the spread of diseases and increase farming output by precisely forecasting the intensity levels of various wheat diseases. This study’s use of federated learning emphasizes the value of anonymity in machine learning and the possibility of common learning methods to create accurate and effective machine learning models. The findings of this research show how cooperative learning CNN can be used to evaluate the degrees of seriousness of wheat diseases, as well as to increase farming output and stop the spread of diseases.