An Empirical Study of ResNet50 Hyperparameter Tuning for Plant Disease Classification

Deep learning techniques have gained popularity in identifying plant diseases owing to their potential for early and accurate identification of diseases. This study presents an empirical investigation into optimizing the hyperparameters of the ResNet50 architecture for plant disease classification. The work conceived explores the optimal configuration in model depth, dense layer configurations, learning rates and the inclusion of batch normalization to discern their impact on classification accuracy. A systematic empirical analysis reveals insights into the interaction of hyperparameters and model performance. The study highlights the configurations that enhance ResNet50's ability to classify plant diseases accurately. This paper provides detailed insights into improving ResNet50 for plant disease classification by exploring hyperparameter tuning without directly comparing several models. The results provide useful recommendations to practitioners for improving the efficiency of deep learning models in diagnosing agricultural diseases.