Lotus Disease Diagnosis Using Combined CNN and SVM with Max Pooling and Convolutional Layers
Deepak Banerjee, Vinay Kukreja, Rishika Yadav, Kireet Joshi, Amitoj Singh
Abstract
This study presents an evaluation of a classification model employing convolutional neural network (CNN) and support vector machine (SVM) algorithms for the identification of various diseases in lotus plants. The model's performance is assessed using precision, recall, F1-score, support, and accuracy metrics for each disease class. The results demonstrate the model's effectiveness in accurately distinguishing between different diseases. For instance, Leaf Spot achieves a precision of 93.97%, recall of 91.60%, F1-score of 92.77 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> , and support of 595 instances, with an overall accuracy of 97%. Similar high-performance metrics are observed for Rust, Pythium Root Rot, Leaf Blight, Stem Rot, and Lotus Mosaic Virus classes, showcasing the model's robustness. The weighted average metrics, considering class distribution, yield precision, recall, and F1-score values around 92.86%. The macro average indicates an overall average performance across all classes, while the micro average reveals a precision and recall of 92.85%. These findings validate the efficiency and reliability of the CNN and SVM model in accurately identifying and managing lotus leaf diseases, thus contributing to crop health monitoring and disease management in lotus plantations.