Sri Lankan leafy tea quality grades classification using deep learning

Abstract

Tea is a popular beverage in almost every corner of the world because of its unique taste and aroma. Tea originates from the leaves of the Camellia sinensis plant. Tea grading involves analyzing different parameters to identify the quality grades of tea, including the size and appearance of the tea leaf, colour, aroma, flavour, texture, uniformity and liquor. Manual methods and Sifter machines are used for the current grading process. Colour sorter machines classify tea grades by analyzing the colour of tea leaves and sifter machines grade tea leaves by analyzing the size of tea particles. Current tea grading methods are time-consuming, laborious, expensive and sometimes inaccurate. To address these challenges, machine learning approaches have been introduced as effective solutions. In the past few years, many researchers introduced different machine learning algorithms to classify tea grades, including artificial neural network (ANN), K-nearest neighbors (KNN) and Support Vector Machine (SVM) [1]. Deep learning provides powerful algorithms for computer vision tasks including image classifications. This study aims to propose a novel approach for the Sri Lankan tea grading process using deep learning as AI is still rarely used in this domain.