Improved <scp>CycleGAN</scp> Based Quality Level Identification of Tea

ABSTRACT Chinese tea culture has a long history, and tea tasting is a relaxation and a kind of art philosophy. Thus, they have a set of quality evaluation system since ancient times. There are many varieties and brands of Chinese tea. It is relatively difficult to identify tea quality without a professional analyzer. To this end, we propose a quality identification method by computer vision based on deep learning methods. The method can help to analyze the complex characteristics of tea. The results show that the proposed system is capable of distinguishing the Longjing green tea grades accurately. Our findings will assist people to achieve intelligent inspecting and grading of tea by vision.