This is the authors' abstract. We don't add key points for this paper.
To address the issues of low small-object detection accuracy and high computational cost in existing steel surface defect algorithms, this paper proposes CAP-YOLOv11n, an enhanced model based on YOLOv11n. Static local and dynamic global aggregation matrices are introduced before the detection head to fuse contextual information, strengthening key feature representations and suppressing background noise. An adaptive threshold mechanism combined with dynamic loss adjustment mitigates positive-negative sample imbalance and improves smallobject detection. Additionally, partial convolution is incorporated into the C3K2 module, applying convolution selectively to key channels, thus reducing computational redundancy. Experiments show CAP-YOLOv11n achieves an average precision of 81.6 %, a <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">5. 4 %</tex> improvement over the baseline, with significantly lower computational costs.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2025) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: