Hybrid Vision Transformer Method for Detecting and Classifying Coconut Leaf Diseases

The sustainability of coconut farming is significantly challenged by leaf diseases, which cause substantial crop losses globally. Accurate detection of these diseases is essential for effective management. This work introduces a novel hybrid model that combines Convolutional Neural Networks (CNN) with Vision Transformers (ViT) to leverage the complementary strengths of local feature extraction and global attention mechanisms. By integrating these architectures, the hybrid model provides a more comprehensive feature representation, enabling improved detection and classification of five common coconut leaf diseases: yellowing, drying, flaccidity, leaflet abnormalities, and caterpillar infestations. Utilizing a curated dataset the model achieved an accuracy of 91.6%. This surpasses traditional CNN and K-Nearest Neighbors (KNN) models in both accuracy and efficiency, offering significant advantages in identifying subtle disease manifestations. The novelty of this hybrid approach lies in its ability to capture intricate local patterns while maintaining global contextual relationships, making it highly adaptable for complex agricultural datasets. Compared to standalone methods, the hybrid model demonstrates superior robustness, scalability, and precision, making it suitable for real-world applications, such as portable and cost-effective disease detection systems. Furthermore, its adaptability to diverse environmental conditions and potential for early-stage detection underscores its significance in advancing sustainable agriculture. By addressing critical gaps in coconut disease management, this work provides a scalable solution for improving crop health and enhancing farmer livelihoods.