As one of the most commonly used materials

steel plays a significant role in industrial applications

particularly in steel surface defect detection.In order to improve the accuracy and speed of steel surface defect detection

an improved YOLO-V7-based steel defect detection algorithm is proposed. This algorithm utilizes a GhostConvmodule to reduce the parameter redundancy in the training. It also incorporates a SimAM attention mechanism into the pooling convolution module

enabling the model to generate three-dimensional attention weights and enhance the learning capability of features. Additionally

a Global Attention Mechanism is integrated into the pooling convolution module to allow the network to focus on more information in the training.Furthermore

the dataset is clusteringby using K-means++ to obtain new anchor boxes

thereby improving the accuracy and robustness of the algorithm model. Experimental results demonstrate that the proposed algorithm achieves a 5.8% increase in mAP comparing with the YOLO-V7 algorithm

with an FPS value of 59.7. The improved algorithm effectively enhances the accuracy of steel defect detection and is better suited for industrial inspection environments.