Classification of Tea Leaf Diseases Based on ResNet-50 and Inception V3

Abstract

Technological advances have made a major contribution to controlling plant diseases. One method for resolving issues with plant disease identification is the use of deep learning for digital image processing. Tea leaf disease is a plant disease that requires fast and effective control. So, in this study, we adopted the Convolutional Neural Network (CNN) architectures, namely ResNet-50 and Inception V3, to classify six types of diseases that attack leaves. The amount of data used was 5867, which were divided into six classes, namely healthy leaf, algal spot, brown blight, gray blight, helopeltis, and red spot. The process of distributing the data involves randomly splitting it into three portions, with an allocation of 80% for training, 10% for validation, and 10% for testing. The process of classification is carried out by adjusting the use of batch sizes in the training process to maximizehyperparameters. The batch sizes used are 16, 32, and 64. Using three different batch size scenarios for each model, it shows that ResNet-50 has better performance on batch size 32 with an accuracy value of 97.44%, while Inception V3 has the best performance on batch size 64 with an accuracy of 97.62%..