Adoption Tea Leaf Disease Detection Using ResNet50 and Advance Hyperparameter

Advanced technology has become a key factor in managing plant diseases. One effective approach to diagnosing plant diseases is the use of deep learning models within the field of image processing. Tea leaf disease is one such condition that demands rapid and efficient intervention. This study employs a Convolutional Neural Network (CNN) architecture, specifically ResNet-50 and Inception V3, to classify six different diseases affecting tea leaves. To enhance the performance of ResNet-50 and Inception V3, data augmentation techniques were incorporated into the training process to increase data variability, and hyperparameters were fine-tuned to improve model effectiveness. The research includes four experimental scenarios aimed at achieving optimal outcomes. Three of these scenarios involve adjusting hyperparameters such as the optimizer, learning rate, and batch size. Classification was performed using three sets of data augmentation methods: zoom, rotation, and flip; zoom, rotation, and shear; and zoom, rotation, and shift. Fine-tuning hyperparameters and applying the right data augmentation techniques significantly improved ResNet-50's performance. The final results revealed that ResNet-50 outperformed the other models in classifying tea leaf diseases, achieving an average accuracy of 98.30% with the Adam optimizer, a learning rate of 0.0001, a batch size of 16, and the augmentation techniques of rotation, zoom, and flip.