ATTENTION ENHANCED HYBRID VGG19 CNN MODEL FOR ROBUST SUGARCANE DISEASE CLASSIFICATION

Abstract

Accurate and timely identification of sugarcane diseases is essential for improving crop productivity and supporting sustainable agricultural practices. This study proposes an attention-enhanced hybrid deep learning framework that integrates VGG19-based transfer learning with custom convolutional neural network (CNN) layers for robust disease classification. To address the challenge of high inter-class similarity among sugarcane diseases, a lightweight attention mechanism is incorporated to selectively emphasize disease-relevant regions and enhance discriminative feature learning. A dataset comprising 2000 labeled sugarcane leaf images across five disease classes—Cercospora Leaf Spot, Helminthosporium Leaf Disease, Rust, Red Dot, and Yellow Leaf Disease—was used for training and evaluation. The proposed model was assessed using standard performance metrics, including accuracy, precision, recall, and F1-score. Experimental results demonstrate that the proposed approach achieves an accuracy of 94.6%, outperforming baseline CNN and conventional VGG19 models. Furthermore, the model is benchmarked against state-of-the-art architectures such as ResNet50, EfficientNet-B0, and MobileNetV2, demonstrating superior performance in feature discrimination and classification robustness. The integration of attention-based feature refinement with transfer learning significantly improves localization of disease-specific patterns and reduces misclassification among visually similar classes. The proposed framework offers a scalable and efficient solution for real-time crop disease detection and precision agriculture applications. Future work will focus on optimizing the model for edge deployment and extending the framework to multi-crop disease classification