DefectDet: An Annotated Dataset of Printed Circuit Boards with Surface Defects

Read full text
Authors: Ashraful Azam, Masum Mushfiq Ishti, Khondakar Muntaseer Shams, Ahmed Jawad Rashid, Nadim Ahmed
Cited by: 0

Abstract

The following dataset provides a collection of 268 annotated high-resolution images of manufactured double-layer PCBs, produced via standardised manufacturing protocols. Surface defects were intentionally introduced during the design phase through precise vector-level modifications of SVG and Gerber files and were physically realised during manufacturing. Five common surface defect categories are included: short circuit, open circuit, spurious copper, spur, and missing pad. Image resolutions range from approximately 1000 × 1000 to 2400 × 2400 pixels, providing full-board coverage for each PCB. Detailed bounding box annotations were generated for all visible defect instances using the Roboflow tool, resulting in approximately 1300 annotations with an average of 5 defect instances per image. All annotations follow the COCO (Common Objects in Context) JSON format.

Fields of Study & Subjects

Industrial Vision Systems and Defect Detection Infrastructure Maintenance and Monitoring Advanced Neural Network Applications

Publication details

Journal / source
Mendeley Data
Publication date
15 February 2026
Language
EN
Document type
Dataset