Related Experiment Video
Updated: Aug 6, 2026
11:34
Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography
Published on: May 15, 2017
RFDD: A High-Quality Dataset based on Metrology for Fastener Defect Detection in High-Speed Railways
Bin Wang1,2,3, Chenbo Pei1,2,3, Xingchuang Xiong1,2,3
1Center for Metrology Scientific Data, National Institute of Metrology, Beijing, 100029, China.
Scientific Data
|July 17, 2026
Summary
The Railway Fastener Defect Dataset (RFDD) offers a realistic benchmark for intelligent railway inspection. It features full-scene images with precise defect annotations, improving computer vision for infrastructure maintenance.
Area of Science:
- Computer Vision
- Railway Engineering
- Infrastructure Monitoring
Background:
- Existing datasets lack realistic railway inspection scenarios.
- Current methods struggle with diverse fastener defect detection.
- Need for metrology-oriented datasets in railway maintenance.
Purpose of the Study:
- Introduce the Railway Fastener Defect Dataset (RFDD).
- Provide a high-quality benchmark for intelligent railway inspection.
- Establish engineering-oriented defect definitions for operational criteria.
Main Methods:
- Collected 1,350 high-resolution full-scene images.
- Annotated over 8,100 fastener instances with pixel-level semantic segmentation.
- Ensured metrological consistency via six visual principles (geometry, optics, etc.).
Main Results:
- RFDD includes Normal class and five defect categories (Missing, Inverted, Displaced, Deformed, Fractured).
- Dataset preserves original multi-fastener spatial layouts.
- Validated using state-of-the-art detection architectures.
Conclusions:
- RFDD serves as a realistic, reliable, and challenging benchmark.
- Enables evaluation of advanced computer vision algorithms for railway inspection.
- Facilitates development of intelligent sensing systems for infrastructure maintenance.
