Strawberry Leaf Disease Severity Decoded for Agriculture: A Federated Learning CNN Approach

In this research paper, agricultural diseases can be detected using innovative technologies. To discuss classification approaches to strawberry leaf disease using a Federated Learning (FL) framework and Convolutional Neural Networks (CNN). A study assessing leaf disease severity in four grades used datasets from four customers. Using FL in this context enables local data analysis from different sources. To protect privacy and keep all training on-site, averages are federated to create a comprehensive global model. To found that the FL-CNN model can accurately grade strawberry leaf disease severity. Four clients performed well in precision, recall, F1 score, and overall accuracy in an analysis of four clients. The model results could be understood more nuancedly by calculating macro-, micro-, and average-weighted results. As for the standardized performance metrics across clients, including macro averages, they were 85.19%, 82.06 %, 86.61 %, 83.10 %, etc. It has consistently emphasized the stability of its disease severity classification since its founding. Weighted average values, which consider class imbalance, also indicated the model's strength, with 84.95% and scores between 81.5% and 79.0%. These microaverages were 84.97 percent, 81.60 percent, and ovmaxacing respectively. The study's design, which involves converting local data into a one-stop global model based on federated averaging, hints at how FL has a tremendous advantage for agricultural applications (compared to centralized learning) in situations where privacy and distributed training are crucial. To used this technique successfully in our study to develop more precise and refined models for disease detection, which improved measured results on disease detection, which are crucial for effective crop management and implementing strategies for controlling or preventing any given disease.