Fresh Tea Leaves Classification Using Inception-V3

Tea is a kind of favorite commodity all over the world. In the process of tea production, the classification of different kinds of fresh tea leaves is very important. In this paper, based on the Inception-v3 model Keras platform, we used transfer learning technology to conduct experiments on the fresh tea leaves dataset created by ourselves. In combination with the Inception-v3's inherent performance, saving parameters, accelerating computation, and reducing over-fitting, which can bring convenience in practical production applications. Through the experiment, the discrimination accuracy rate of fresh tea leaves can reach more than 95%. After that, through the data augmentation technology, we can make the discrimination accuracy rate of fresh tea leaves reach 98%, which brings good news to industrial production.