Answers are grounded in this paper and cite their sources.
Proposed model enhances defect detection accuracy in steel strips, suggesting improved efficiency via shared convolution.
To address the challenges of suboptimal detection accuracy, high model complexity, and imbalanced performance in detection accuracy, computational speed, and model complexity for existing models in steel strip surface defect detection, this paper proposes MLS-YOLOv11, a steel strip surface defect detection model based on multilayer feature fusion and shared convolutions. First, to capture more comprehensive defect features and enhance detection performance, a multilayer feature fusion diffusion pyramid network is designed to replace the original neck network. Second, a lightweight shared detection head is proposed, which reduces computational costs while improving detection accuracy through shared convolutions and a Scale layer incorporating learnable dynamic factors. Finally, the Shape-IoU loss function replaces the original loss function in YOLOv11 to enhance bounding box regression precision. Experimental results demonstrate that compared to the YOLOv11 algorithm, MLS-YOLOv11 achieves mean average precision improvements of 6.8%, 1.8%, and 1.6% on the NEU-DET strip steel defect dataset, the HIT-UAV small object dataset, and the GC10-DET dataset, respectively, while reducing parameter count and computational load by approximately 7.7% and 3.2%. The detection speed reaches 127 fps. The improved model effectively enhances detection accuracy for steel strip surface defects, reduces model complexity, and achieves an optimal balance among detection accuracy, computational speed, and model complexity. This provides a novel solution for quality control and production management of steel strips.
No takes yet. Share an insight, caveat, or question.
Lei et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: