Utilizing Federated Learning and CNNs for Severity Analysis of Fenugreek Leaf Diseases

To effectively control fenugreek leaf diseases, which represent a severe danger to crop yield, it is necessary to have precise and fast detection systems. This study suggested a unique method for identifying and categorising fenugreek leaf illnesses according to severity levels, combining federated learning with convolutional neural networks (CNN). This study set out to develop a privacy-preserving, decentralized learning model that could assess the severity of fenugreek leaf diseases across various customers' information, enabling more robust and generic predictions. The effectiveness of the local models was evaluated using data from five customers, and the results showed outstanding precision, recall, F1-score, and accuracy values ranging from 79.48% to 97.41%, 82.51% to 97.35%, and 0.95 to 0.99, respectively. Then, a global model with precision, recall, F1-score, and accuracy ranging from 91.14% to 94.05%, 91.70% to 94.49%, 91.25% to 94.19%, and 0.97 to 0.98, respectively, were created using the federated learning method. The macro-average values varied from 91.36% to 94.24% in evaluating the three distinct averaging methods—macro-average, weighted-average, and micro-average—reflecting a dependable and well-rounded performance across all severity levels. The weighted-average figures showed high performance, especially for most courses, and varied from 91.99% to 94.71%. The micro-average statistics, which consider every instance equally, ranged from 91.89% to 94.66%, demonstrating solid performance in every case. As a result, the study showed how federated learning combined with CNN might effectively be used to identify and classify fenugreek leaf diseases, attaining good performance across various measures. This research expands the potential of AI-driven agricultural practices and helps create more efficient and productive farming systems.