Integration of YOLOv11 and Convolutional Neural Network in a Deep Learning Approach for Coffee Bean Defect Detection and Classification
Year: 2026
Type: conference-paper
Abstract: The coffee industry is a strategic commodity that significantly contributes to the global and national economy. Coffee bean quality strongly influences flavor and market value, while defective beans—such as broken, moldy, or quaker beans—can reduce overall quality. Manual sorting methods, still widely used by farmers and small-scale producers, are time-consuming, inefficient, and prone to human error. This study proposes an automated deep learning–based system for detecting and classifying defective coffee beans by integrating YOLO for object detection and EfficientNetV2 as the classifier. A dataset of 5,636 coffee bean images from multiple sources was used. The system was evaluated through black box testing to ensure the functionality of the web interface and performance testing using a confusion matrix. Results show that YOLOv11 achieved an [email protected] of 98.83%, while EfficientNetV2 obtained a test accuracy of 93.81%. The proposed system demonstrates strong potential to improve coffee sorting by providing a faster, more accurate, and efficient alternative to manual methods.
Institution: Universitas Halu Oleo
Language: English
Subfield: Industrial and Manufacturing Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7127048386