SD-YOLO: A lightweight manhole cover defect detection method with diffusion model data augmentation

doi:10.1016/j.rineng.2026.112594

SD-YOLO: A lightweight manhole cover defect detection method with diffusion model data augmentation

2026-08-22

0

OA

AI

DOI：10.1016/j.rineng.2026.112594

Original

Original request for help

Save

Abstract

En 中文

• A lightweight defect detector SD-YOLO is proposed for manhole cover inspection. • A stable diffusion-based augmentation strategy is used to enrich defect data. • A Local-Global Feature Attention enhances defect texture and structure modeling. • A new dataset with diverse defect types and real-world scenarios is constructed. • SD-YOLO achieves 83.9% F1-score, showing strong accuracy and robustness.

Keywords:

Manhole cover defect detection

Fine-grained

Data augmentation

Local-global attention

SD-YOLO: A lightweight manhole cover defect detection method with diffusion model data augmentation

Abstract

Journal

Organization