PCB-Defect: An Annotated Dataset for Surface Defect Detection in Printed Circuit Boards

The PCB Defect Detection dataset comprises 230 images of uniquely manufactured single-layer printed circuit boards (PCBs) with authentic, chemically induced defects. Each PCB was fabricated using a controlled chemical etching process, ensuring that the defects closely replicate those encountered in industrial production environments. The dataset captures six key defect types: missing pad, mouse bite, open circuit, short circuit, spur, and spurious copper.
Each image in the dataset has been annotated with fine-grained bounding boxes localizing all visible defect instances, resulting in an average of 7.4 annotations per image and a total of 1,704 defect annotations across the six defect categories. The image resolutions vary ranging from 2.57 to 31.26 megapixels (800×600 to 6000×4000 pixels). Annotations are provided in COCO-style JSON format, which ensures seamless compatibility with machine learning object detection frameworks.