Sewer Defect Detection Dataset
收藏资源简介:
SEDD(Sewer Defect Detection Dataset)是一个面向工业与市政场景的管道表面缺陷检测数据集,旨在支持和评估真实环境下的缺陷检测算法研究。该数据集共包含 7,563 张高分辨率图像,分辨率为 1920 × 1080,图像来源于实际下水管道巡检过程。SEDD 覆盖了三类典型的管道表面缺陷类型:裂缝(crack)、树根侵入(root)和沉积物堆积(deposit),同时也包含大量无缺陷样本。其显著特点在于缺陷样本与无缺陷样本分布更贴近真实应用场景,其中无缺陷图像占比接近三分之二,这对检测模型在低缺陷比例、强类别不平衡条件下的鲁棒性提出了更高要求。SEDD 数据集适用于下水管道自动巡检、表面缺陷检测与工业视觉质量控制等研究方向,可从百度网盘下载获取。
SEDD (Sewer Defect Detection Dataset) is a pipeline surface defect detection dataset designed for industrial and municipal scenarios, aiming to support and evaluate research on defect detection algorithms in real-world environments. This dataset contains a total of 7,563 high-resolution images with a resolution of 1920 × 1080, which are collected from actual sewer pipeline inspection operations. SEDD covers three typical types of pipeline surface defects: crack, root intrusion, and deposit, and also includes a large number of defect-free samples. Its prominent feature is that the distribution of defective and defect-free samples is more consistent with real-world application scenarios, where defect-free images account for nearly two-thirds of the total, which imposes higher requirements on the robustness of detection models under conditions of low defect proportion and severe class imbalance. The SEDD dataset is suitable for research directions such as automatic sewer pipeline inspection, surface defect detection, and industrial visual quality control, and can be downloaded via Baidu Netdisk.




