Development of an Improved YOLOv7-Based Model for Detecting Defects on Strip Steel Surfaces

The detection of defects on the surface is great importance for both production and application strip steel. In order to detect accurately, an improved YOLOv7-based model detecting steel developed. To enhances ability extract features identify small features, ConvNeXt module introduced backbone network structure, attention mechanism embedded in pooling module. reduce size improves inference speed model, C3 was used replace ELAN head. experimental results show that, compared with original models, mAP proposed reached 82.9% by 6.6%. can satisfy need accurate identification defects.