Fusion Lightweight Steel Surface Defect Detection Algorithm Based on Improved Deep Learning

In industrial production, timely and accurate detection and identification of surface defects in steel materials were crucial for ensuring product quality, enhancing production efficiency, and reducing production costs. This study addressed the problem of surface defect detection in steel materials by proposing an algorithm based on an improved version of YOLOv5. The algorithm achieved lightweight and high efficiency by incorporating the MobileNet series network. Experimental results demonstrated that the improved algorithm significantly reduced inference time and model file size while maintaining performance. Specifically, the YOLOv5-MobileNet-Small model exhibited slightly lower performance but excelled in inference time and model file size. On the other hand, the YOLOv5-MobileNet-Large model achieved a slight performance improvement while significantly reducing inference time and model file size. These results indicated that the improved algorithm could achieve lightweighting while maintaining performance, showing promising applications in steel surface defect detection tasks. It provided an efficient and feasible solution for this important domain, offering new insights and methods for similar surface defect detection problems and contributing to research and applications in related fields.

@article{Ren2024,
  title     = {Fusion Lightweight Steel Surface Defect Detection Algorithm Based on Improved Deep Learning},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {15},
  number    = {5},
  year      = {2024},
  publisher = {The Science and Information Organization},
  author    = {Fei Ren and Jiajie Fei and HongSheng Li and Bonifacio T. Doma Jr},
  doi       = {10.14569/IJACSA.2024.0150537},
  url       = {https://doi.org/10.14569/IJACSA.2024.0150537}
}