Akademik Veri Yönetim Sistemi
Kuzey Ege Teknik Bilimler ve Teknoloji Dergisi, cilt.2, sa.1, ss.33-50, 2025 (Hakemli Dergi)
Printed Circuit Board (PCB) defect detection is critical in electronics manufacturing, as undetected faults can lead to severe quality control issues. Recent advancements in deep learning, particularly object detection models, have significantly improved inspection systems' accuracy and speed. This study explores the performance of the YOLO11 (You Only Look Once version 11) object detection architecture on a multi-class PCB defect dataset. Five YOLO11 variants—YOLO11n, YOLO11s, YOLO11m, YOLO11l, and YOLO11x—were trained and evaluated under identical conditions using high-resolution images containing six defect types. Metrics such as mAP@50, mAP@50-95, and FPS were used for evaluation. Results demonstrate that YOLO11l achieved the highest mAP@50-95 of 0.551, while YOLO11n achieved up to 166 Frame Per Second (FPS) on an NVIDIA A100 GPU, confirming its real-time capability. Comparative analysis against state-of-the-art models confirms that YOLO11 variants offer an effective trade-off between accuracy and efficiency. This study positions YOLO11 as a strong candidate for real-time PCB inspection systems.