Tiny Machine Learning for Classifying Specialty Coffees

Isabela V. De Carvalho Motta, Felipe Augusto de Figueiredo, Samuel Mafra

DOI: 10.14209/sbrt.2024.1571028147

Evento: XLII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2024)

Keywords: IoT TinyML Precision Agriculture Machine Learning

Abstract

The consumption of specialty coffee has increased around the world. Specialty coffee is free from impurities and defects. Specialty coffee is produced in smaller quantities, and its production process is hard and expensive. Traditional coffee beans have defects that affect the flavor of the coffee. It is essential to select the type of coffee to guarantee the best cost and quality. When a human manually classifies the type of coffee, there may be interference due to the human condition. This process has the disadvantage of being subjective. Few studies used Machine Learning methods for predicting specialty coffee classification by analyzing images. This article proposes a framework for classifying specialty coffees by applying Tiny Machine Learning techniques. We developed a model that can help accurately analyze and classify the coffee process and inform the quality of coffee without human interference. We achieved 100\% accuracy, and we believe our model can be used effectively in the coffee industry.

Download