Weld-Crack Detection Using YOLOv6-NW
Author Information
Author(s): Wang Yiming, Shang Lunhua, Li Bin, Liu Yu, Ji Ye, Hao Linbo, Zhang Yang, Li Yuchen, Tian Menghan, Cheepu Murali Mohan
Primary Institution: Luoyang Institute of Science and Technology
Hypothesis
Can a lightweight detection network improve weld-crack detection accuracy and efficiency?
Conclusion
The YOLOv6-NW model significantly outperforms YOLOv5 and maintains high detection accuracy while being lightweight.
Supporting Evidence
- YOLOv6-NW has only 16% of the parameters of YOLOv6-N but maintains comparable accuracy.
- The model's precision and recall remain above 0.9 under low-resolution conditions.
- Data augmentation techniques were used to enhance the diversity of the training dataset.
- The model shows strong robustness to variations in illumination during detection.
Takeaway
This study created a smart system to find cracks in welds better and faster, using special techniques to make the system work well even with fewer examples.
Methodology
The study used data augmentation techniques and developed a lightweight detection model called YOLOv6-NW for weld-crack detection.
Limitations
The study may be limited by the artificial nature of the weld-crack samples and the specific conditions under which the experiments were conducted.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website