Android application determines the quality of coffee roasting by photo

(ORDO NEWS) — Scientists from Thailand have developed an Android application that, based on a neural network, determines the quality of roasted coffee beans.

The neural network was trained on 4800 images of coffee beans of different roast levels, and now determines the quality of the beans no worse than an experienced barista.

Coffee beans are the seeds of the Coffea plant, which is grown in several parts of Central and South America, as well as Africa, the Middle East, and Asia.

The quality and taste of coffee depends on various factors, including the conditions in which Coffea plants are grown, the storage, processing and roasting of coffee beans.

It is not always easy for a person to determine the degree of roasting of coffee beans, because sometimes this requires special training or experience.

Researchers at the King Mongkut University of Technology in Thonburi in Thailand have developed a smartphone app that can determine the roast level of a batch of coffee beans by simply analyzing their images.

This application, presented in the article by scientists , is based on deep learning methods.

Android application determines the quality of coffee roasting by photo 2

The deep learning model is based on a convolutional neural network (CNN). The researchers trained it on a dataset containing images of four different varieties of coffee beans. The dataset contained a total of 4800 photographs, 1200 for each variety.

The researchers’ deep learning method is based on analyzing the color of coffee beans. The Android application allows users to quickly determine the roast level of a particular batch of beans by simply sending their image to the input of the neural network.

The researchers’ approach has already shown promising results. However, their neural network does not take into account the origin of coffee beans, which can also affect their color, and this sometimes leads to errors.

In their next studies, the researchers hope to increase the accuracy of the neural network, but for this they will need a more diverse data set, writes Techxplore .


Contact us: [email protected]

Our Standards, Terms of Use: Standard Terms And Conditions.