Related Experiment Video
Updated: Jul 1, 2026
11:34
Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography
Published on: May 15, 2017
Industrial Printed Circuit Board Surface Defect Dataset for Object Detection
Hao Yan1,2, Xiaoguang Yu3,4, Bifang Ma1
1School of Electronic and Mechanical Engineering,Key Laboratory of Nondestructive Testing, Fujian Polytechnic Normal University, Fuqing, 350300, Fujian, China.
Scientific Data
|June 29, 2026
Summary
This study introduces PCB-IND, a real-world dataset for printed circuit board (PCB) surface defect detection. It features industrial images to improve automated optical inspection (AOI) model performance in manufacturing.
Area of Science:
- Materials Science
- Computer Vision
- Manufacturing Engineering
Background:
- Printed circuit board (PCB) defect detection is crucial for electronic manufacturing quality control.
- Existing datasets lack real-world industrial conditions, including varied illumination and defect types.
- Current methods struggle with the complexity of defects found in actual production lines.
Purpose of the Study:
- To introduce PCB-IND, a large-scale, real-world dataset for PCB surface defect detection.
- To provide a benchmark dataset that captures industrial complexities like non-uniform illumination and diverse defect scales.
- To facilitate the development and evaluation of robust defect detection models for industrial applications.
Main Methods:
- Collected 4,789 real-world images from an industrial automated optical inspection (AOI) system.
- Annotated 5,932 defect instances across eight typical surface defect categories.
- Evaluated dataset usability with representative object detection models.
Main Results:
- PCB-IND dataset preserves industrial imaging characteristics: non-uniform illumination, high contrast, and cross-scale defects.
- The dataset exhibits a natural long-tail distribution of defects, reflecting real production scenarios.
- Experimental results demonstrate the dataset's suitability for training and evaluating object detection models under industrial conditions.
Conclusions:
- PCB-IND offers a valuable real-world data resource for advancing industrial defect detection research.
- The dataset supports research in early-stage process monitoring and defect repair decision-making.
- It enables the development of more accurate and reliable automated optical inspection systems for PCB manufacturing.
