An Empirical Study on Different Deep Learning Models for Disease Detection in Tea Leaves

For disease identification in plants, convolutional neural networks can be engaged. CNNs perform involuntary feature mining along with classification. Here, we carry out an empirical study on five CNN models for detecting diseases in tea leaves. The models used were VGG19, ResNet152V2, InceptionV3, MobileNetV2, and DenseNet201. Transfer learning models were used for the study. The models were retrained on diseased tea leaf image dataset. Model performances were monitored to find the best of all. The experiments indicate DenseNet201 obtains the highest detection accuracy among all the models used. DenseNet201 was then employed to detect diseases for apple, tomato, and potato datasets.