A Defect Detection Approach for Hot-Rolled Steel Using an Improved YOLOv8

Accurate detection of steel strip surface defects is essential for ensuring product quality and production efficiency. To address the limitations of conventional methods, this paper proposes an enhanced lightweight deep learning model based on the YOLOv8n framework. The model introduces an EMAtt-C2f module to improve feature extraction, a receptive field enhancement module for better recognition of complex defects, and optimizes the loss function using MPDIoU for improved bounding box accuracy. Experimental results on the NEU-DET dataset show an mAP50 of 80.1%, a 6.1% improvement over YOLOv8n, with minimal increase in parameters and computational complexity. This approach provides a practical and efficient solution for steel strip defect detection in industrial applications.