Deep Learning based Coffee Beans Quality Screening

• Published in: 2022 IEEE International Conference on e-Business Engineering (ICEBE) pp. 271 - 275
• Main Authors: Shao, Bing, Hou, Yichen, Huang, Nianqing, Wang, Wei, Lu, Xin, Jing, Yanguo
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2022
• Subjects: Coffee bean screening; Coffee beans; coffee beans screening; Coffee beverages; Companies; Computer vision; Consumer market; Convolutional neural network; Convolutional neural networks; Cost effective; Cost effectiveness; Deep learning; Insect damage; Insects; Production and Automation Engineering; Produktion och automatiseringsteknik; Screening system; Screening tests; Virtual Production Development (VPD)
• ISBN: 9781665492454; 1665492457; 9781665492447; 1665492449
• DOI: 10.1109/ICEBE55470.2022.00054

Coffee bean quality screening is a time-consuming work, and its workload increases abruptly with the rapid development of coffee beverage consumer market. In this work, a CNN-based classifier is developed to categorizing the coffee beans into sour, black, broken, moldy, shell, insect damage and good beans. The screening test results show that the screening accuracy could reach more than 90% for all other beans except for shell beans (88%). Therefore, the proposed method is feasible and promising. Moreover, a cost-effective automatic coffee bean screening system using the developed classifier is manufactured and implemented for a local company.