Green Coffee Bean Quality Inspection Using Computer Vision Systems

Abstract

In the 19th century, the Philippines led global coffee exports until the 1890s, boasting ideal conditions for growing Arabica, Robusta, Excelsa, and Liberica beans. However, traditional methods of quality inspection were slow and errorprone, impacting industry reputation and profitability. This study introduces a method for green coffee bean grading and quality inspection using computer vision technology. Specifically, the study presents a software designed to improve the accuracy and efficiency of bean quality assessment. The software was trained on over 9000 coffee beans, with 65% of the data allocated for training, 25% for validation, and 15% for testing. The YOLOv8n Final Model achieved over 90% accuracy within 75 epochs. ACER NITRO 5 and Lenovo IdeaPad Gaming laptops, equipped with Nvidia graphics, delivered stable visual performance at 28-33 frames per second (FPS). The 48MP microscope camera provided clear and reliable images during sampling and testing, ensuring precise and dependable quality evaluations.