Improved Real Time Printed Circuit Board Fault Detection

Ensuring the quality of PCB products is vital, with various inspection methods and technologies being proposed. Conventional methods for visually examining PCBs present various constraints and are frequently unable to guarantee flawless outcomes. Machine vision technologies for automatic inspection of PCB surfaces are now powerful and effective tools for addressing quality control issues in the industrial field. Hence, this research focuses on identifying surface flaws in PCB images using a lightweight YOLOv8 model (YOLOv8n-tiny). In summary this study mainly focuses on enhancing real-time object detection model performance capable of accurately identifying and locating fault within PCBs under real time situation using a dataset containing six categories of PCB defects. Furthermore, we carried out a comparison study between our proposed model and other models mentioned in the related work section. Our detector outperformed the existing models by achieving a mAP@50 of 99.08%.