A Hybrid Model for Detection of Grey Blight Disease Using Deep Learning

This paper makes an attempt to control ‘grey blight’ fungal disease through deep learning-based approach for early detection of grey blight in tea leaves, focusing on a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) architecture to classify tea leaves as healthy or diseased. It was achieved through feature extraction capabilities of CNNs to capture spatial characteristic of the leaves such as texture and colour changes simultaneously. LSTM layers are employed to recognize sequential dependencies. The proposed model was trained and evaluated on a dataset of tea leaves collected from various tea gardens in Assam which was augmented to improve model robustness. Our experimental results demonstrate that the hybrid CNN-LSTM model achieves 98.05% accuracy on the test set, outperforming traditional and standalone deep learning models. These findings suggest that the proposed approach could be a viable tool for automated, large-scale tea leaf disease detection, contributing to more efficient disease management and yield improvement.