Investigating the Impact of Optimizers and Epochs on VGG-19 in Tomato Disease Recognition

Deep Learning (DL) models excel in image classification, including identifying crop diseases in Tomato plants facing numerous threats. The performance of the DL model is significantly impacted by dataset size and hyperparameter choices. Optimal performance requires balancing underfitting, overfitting, and selecting an optimizer that suits the dataset and model architecture. This study investigates the impact of different optimizers including Adam, RMSprop, and SGD and varying numbers of epochs such as 8 and 10 on training the VGG19 architecture, using a limited tomato dataset comprising five disease categories sourced from the Plant Village dataset. Results show that the Adam optimizer achieves the highest accuracy of 90%, surpassing RMSprop(85.7%) and SGD(71.6%).Thus, the study highlights the importance of selecting appropriate optimizers and epochs for training DL models, particularly in the context of tomato disease recognition. However, further exploration into different hyperparameters and optimization techniques could potentially enhance model performance even further.