YOLOv8-cop: an improved YOLOv8-based surface defect detection algorithm for special-shaped copper strips

The surface quality of shaped copper strips plays an essential role in the quality of the product. Currently, detecting defects on shaped copper strips primarily relies on manual visual inspection. However, inspectors’ subjective judgment and fatigue may affect the inspection results, leading to poor inspection accuracy and high misdetection rates. This study proposes an algorithm (YOLOv8-cop) to mitigate these issues. Considering the scarcity of datasets, a defect dataset of shaped copper strips was first constructed based on real scenarios, and data augmentation was employed to expand the dataset. Then, the algorithm introduces Adaptive Spatial Feature Fusion (ASFF) to optimize the model detection head, minimizing information loss during multi-layer feature transfer and improving target defect detection ability. A lightweight slim-neck module was used to replace the original YOLOv8’s neck module, thereby reducing computational complexity. Furthermore, the Inner-SIoU was utilized to improve the model’s ability to identify small defects on the surface of shaped copper strips. The proposed algorithm increased the mean Average Precision (mAP) by 4.3% over the original YOLOv8 model. Compared with YOLOv5n, YOLOv7, and YOLOv10n, YOLOv8-cop improved mAP@0.5 by 3.9, 38.0, and 7.5 percentage points, respectively. Meanwhile, this algorithm surpasses the YOLOv8 model in detection speed, satisfying the real-time inspection of shaped copper strip.