Abstract: Steel is susceptible to surface defects caused by processing and environmental factors in modern industrial production, which adversely affect its overall quality and service life. To overcome the shortcomings of existing detection methods, such as inadequate accuracy and high computational complexity, this paper proposes an improved YOLOv8-based algorithm called GSL-YOLO. Specifically, Ghost Bottleneck is introduced in the C2f module to replace the standard Darknet Bottleneck, thereby reducing redundant computation and enhancing feature extraction efficiency. DynamicConv is then adopted in place of the standard CBS module to better adapt to complex scenarios. In order to further improve the detection accuracy for minute defects, SENetV1 is integrated into the neck network, enabling adaptive channel weighting that strengthens attention to key features. In addition, a lightweight shared convolution detection head named LSDECD is designed to address the large parameter volume and computational complexity of existing detection heads, effectively reducing both model parameters and computational overhead. Experimental results demonstrate that, on the NEU-DET steel surface defect dataset, GSL-YOLO achieves a 3 percentage-point increase in mAP50 compared with YOLOv8, while simultaneously reducing the number of parameters by 33% and lowering computational cost by 37%. This work not only improves detection accuracy but also provides a lightweight design that satisfies the requirements of real-time industrial detection.