Enhancing Accuracy of Yellowing Disease Severity Level Detection in Coconut Palms with SVM Regularization and CNN Feature Extraction
Deepak Banerjee, Vinay Kukreja, Satvik Vats, Vishal Jain, Bhawna Goyal
Abstract
This research paper proposes a CNN and SVM approach for detecting and classifying the severity levels of yellowing disease in coconut leaves. The methodology used in this study involves four phases, including Image Retrieval and Normalization, Pattern Recognition and Model Building, Model Discrimination, and Sensitivity Analysis, and Findings and Implications. The model's architecture includes three convolutional layers, three max-pooling layers, and two fully connected layers with regularization. The performance of the model is evaluated using various evaluation metrics, including Precision, Recall, F1-Score, Support, and Accuracy. The model achieves high classification performance in all classes, with F1-Scores ranging from 86.83% to 89.66% and a weighted average accuracy of 88.02%. The model's effectiveness in classifying different severity levels of the disease can be useful for disease diagnosis and management. Further research can be conducted to improve the model's performance and to test its effectiveness on other similar diseases. Overall, the proposed CNN and SVM approach is a promising and effective method for detecting and classifying the severity levels of yellowing disease in coconut leaves.