Single-Layer PCB Surface Defect Detection Using YOLOv5 Deep Learning Model

Printed Circuit Boards (PCBs) are essential components in all electronic systems, and their reliability is crucial for product performance, especially in the consumer electronics sector where single-layered PCBs are commonly used. Detecting surface defects such as missing holes, spurious copper, and open circuits is a critical step in ensuring product quality. This paper proposes an automated surface defect detection system specifically designed for single-layered PCBs using a hybrid deep learning approach. The model integrates YOLOv5, a state-of-the-art object detection algorithm, with the Swin Transformer, a vision transformer known for its hierarchical feature learning and enhanced attention mechanisms. A custom dataset comprising over1000 annotated images from six different single-layered PCBs was collected and labeled for this purpose. The proposed model was trained and evaluated on Google Colab and compared with baseline YOLOv5 performance. Results show a significant improvement in defect detection accuracy, particularly in complex and overlapping defect regions. This work demonstrates the potential of transformer-based models in improving PCB inspection processes and offers a practical solution tailored to the specific requirements of consumer electronics manufacturing.