Grapevine Leaves Classification Using Various CNN Model

Accurate classification of grapevine leaves is crucial for the advancement of vineyard industries, as different grapevine varieties require distinct care techniques. However, the vast array of grapevine varieties presents a significant challenge in achieving precise classification. This study aims to refine grapevine classification models beyond previous research and provide a comprehensive comparison of their results. To achieve this goal, we explore the efficacy of various Convolutional Neural Network (CNN) models, including Standard CNN, DenseNet, and MobileNetV2, leveraging different augmentation techniques and architectures. Through extensive evaluation and comparison, we identify the optimal model for grapevine classification. Our results demonstrate impressive accuracies, with DenseNet and MobileNetV2 both achieving 99%, and Standard CNN achieving 95%. These findings offer a more robust model without data leakage and provide a detailed comparison for each model, refining previous studies and advancing the field of grapevine classification.