Analytical Evaluation of CNN and Capsulenet Architectures for Grape Leaf Disease Prediction

The purpose of this work was to develop a CNN-based deep learning model for enhancing disease detection in Grape leaves. The model was trained on a dataset of approximately 2,400 photographs of grape leaves, containing images of bacterial blight, spider mites and leaf miners.

To ensure model robustness, a k-fold cross-validation approach was implemented for dataset splitting. The developed model exhibited impressive performance in accurately identifying leaf diseases, demonstrating its potential for real-time applications. This study emphasizes the effectiveness of IT-based disease management approaches as complementary strategies to conventional methods, enabling timely interventions and boosting grape sector production. Integrating this deep learning model into agricultural systems allows farmers to benefit from timely and targeted interventions, leading to increased crop yields and economic prosperity.

The utilization of CNN and deep learning techniques in grape farming presents a pathway towards a more environmentally friendly future, showcasing the potential of IT-based methods in revolutionizing disease management. The study reveals that the CapsuleNet model unveils an accuracy rate of 91% in grape leaf disease detection.