PCB DEFECT DETECTION USING DEEP LEARNING: A COMPARATIVE PERFORMANCE ANALYSIS

The rapid expansion of computer, communication and consumer sectors has propelled the production of electronic information devices into a new phase of development. Consequently, the market is seeing a rise in demand for PCBs. PCB defect detection demands high precision and efficiency to identify critical flaws such as spurious copper, mousebites, pinholes, spurs, open circuits and short circuits and play an important role in manufacturing. This paper evaluates YOLOv11’s performance in detecting defects on Printed Circuit Boards (PCBs) and compares it with YOLOv8 and YOLOv10. YOLOv11 achieved a precision of 98.9%, a mean Average Precision (mAP) of 94.7 at IoU0.5 and 79.8% at IoU thresholds 0.5:0.95, with a processing speed of 121 FPS, showcasing its strong performance. Notably, YOLOv11 achieved perfect precision (1.0) for detecting copper defects, underscoring its reliability for specific critical defect classes. While YOLOv10 demonstrated the highest speed due to its Non-Maximum Suppression NMS-free architecture, YOLOv11 balanced accuracy, adaptability, and computational efficiency. These findings affirm YOLOv11’s suitability for real-time PCB defect detection in industrial applications, paving the way for more robust and automated quality inspection systems in smart electronics manufacturing.