Automated Robusta Coffee Quality Grading System Based on ESP32-CAM and Computer Vision Masud Effendi 1*, Usman Effendi 1, Imam Santoso 1, Retno Astuti 1, Wayan Firdaus Mahmudy 2, Adrian Sanjaya 1, Tasya Nadhiva Nur Azmi 1
(1) Department of Agroindustrial Technology, Faculty of Agricultural Technology, Universitas Brawijaya
Veteran Street, Malang, Indonesia
*Email:mas.ud[at]ub.ac.id
(2) Faculty of Computer Science, Universitas Brawijaya
Veteran Street, Malang, Indonesia
Abstract
This research aims to develop an efficient and automated robusta coffee bean quality grading system by utilizing the ESP32-CAM and computer vision. Based on the analysis, the You Only Look Once (YOLOv8) deep learning model with image data augmentation proved to be the best solution, achieving superior performance with an mAP@0.5 value of 0.924 and an F1-score of 0.87. The integration of this model with the ESP32-CAM microcontroller module enables wireless image acquisition of coffee beans, which are then transmitted to a computing device for processing. The YOLOv8 model accurately detects and classifies 26 types of physical coffee bean defects according to the Indonesian National Standard (SNI). The results show that this technological combination provides an accurate, consistent, and affordable quality grading tool, overcoming the limitations of manual methods prone to human error and subjectivity. The implementation of this system has the potential to enhance efficiency and product value for coffee farmers and the industry.
