Optimizing VGG16 Architecture with Bayesian Hyperparameter Tuning for Tomato Leaf Disease Classification

Abstract

This study proposes an optimized VGG16 architecture enhanced through Bayesian Optimization to improve the classification of tomato leaf diseases. The modified model integrates tunable parameters such as dropout rates, convolutional filters, and dense units, while maintaining the foundational structure of VGG16. To further refine performance, Bayesian Optimization is employed to search for the most effective combination of hyperparameters. Experiments conducted using the Tomato Leaf Disease Detection dataset demonstrate that the proposed method outperforms the original VGG16 model, achieving a test accuracy of 97.1% compared to 89.0%. These results underline the importance of architecture customization and systematic hyperparameter tuning for domain-specific deep learning tasks in agriculture.