Related Experiment Video
Updated: Jun 10, 2025
11:34
Subsurface Defect Localization by Structured Heating Using Laser Projected Photothermal Thermography
Published on: May 15, 2017
11.1K
A Lightweight Strip Steel Surface Defect Detection Network Based on Improved YOLOv8.
Yuqun Chu1, Xiaoyan Yu2, Xianwei Rong2
1School of Computer Science and Information Engineering, Harbin Normal University, Harbin 150025, China.
Sensors (Basel, Switzerland)
|October 16, 2024
Summary
This study introduces YOLO-SDS, a lightweight network for strip steel surface defect detection. It significantly improves accuracy and reduces parameters, enhancing quality control in steel production.
Area of Science:
- Materials Science
- Computer Vision
- Artificial Intelligence
Background:
- Strip steel surface defect detection is vital for quality control.
- Existing methods struggle with varying defect sizes and image blur, leading to low accuracy and long detection times.
Purpose of the Study:
- To propose a lightweight strip steel surface defect detection network (YOLO-SDS) based on an improved YOLOv8.
- To enhance detection accuracy and efficiency for strip steel surface defects.
Main Methods:
- Utilized StarNet to replace the YOLOv8 backbone for lightweight optimization.
- Introduced a lightweight module (DWR) and C2f module in the neck for multi-scale feature extraction.
- Incorporated an occlusion-aware attention mechanism (SEAM) in the detection head for complex scenarios.
Main Results:
- The improved YOLO-SDS model reduced parameters by 34.4% compared to the original YOLOv8.
- Achieved a 1.5% increase in average detection accuracy on the NEU-DET dataset.
- Demonstrated good generalization on the deepPCB dataset and outperformed other models in parameter count and detection speed.
Conclusions:
- YOLO-SDS offers significant advantages in parameter count and detection speed for strip steel surface defect detection.
- The integrated modules (StarNet, DWR, SEAM) effectively enhance model performance.
- The proposed model provides an efficient and accurate solution for automated quality control in strip steel production.
