Implementasi YOLOv8 untuk Deteksi Objek Gelas pada Lingkungan Brewing Kopi Menggunakan Dataset Citra Kustom

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The advancement of computer vision technology has accelerated the deployment of object detection systems across various industrial sectors, including visual monitoring in production environments. You Only Look Once (YOLO) is a deep learning-based object detection algorithm recognized for its high inference speed and competitive accuracy on real-time detection tasks. This study implements YOLOv8 for detecting glass objects in a coffee brewing environment using a custom image dataset collected independently. The dataset comprises 152 images with 1 object class annotated using Label Studio in YOLO bounding box format. The YOLOv8n model was trained using SGD optimizer, 640×640 pixel input size, batch size 16, and mosaic data augmentation for 50 epochs. Evaluation used Precision, Recall, F1-score, and mAP@0.5 metrics. Results show mAP@0.5 of 0.847, maximum F1-score of 0.88 at confidence threshold 0.493, and precision of 1.00 at threshold 0.688. Confusion matrix analysis revealed 27 of 36 objects correctly detected with 3 false positives and 6 false negatives. This study demonstrates that YOLOv8 can be effectively implemented for glass object detection in the coffee brewing domain using a limited-scale custom dataset.