NEU-CLS

The NEU-CLS dataset is collected from real industrial production lines. It is publicly released by the Surface Inspection Laboratory of Northeastern University, constructed explicitly for strip surface defect classification. This dataset covers six common types of strip surface defects: Crazing, Scratch, Pitted Surface, Patch, Rolled-in Scale, and Inclusion. The image's defect morphologies are diverse, presenting high complexity and diversity, effectively simulating the challenges of defect detection in practical application scenarios. Each defect category contains 300 images, all of which are grayscale with a resolution of 200×200 pixels.

NEU-CLS数据集采集自真实工业生产线，由东北大学表面检测实验室公开发布，专为带钢表面缺陷分类任务构建。该数据集涵盖6类常见带钢表面缺陷：裂纹（Crazing）、划痕（Scratch）、麻点（Pitted Surface）、斑块（Patch）、压入氧化皮（Rolled-in Scale）以及夹杂物（Inclusion）。数据集内的图像缺陷形态多样，具备高度的复杂性与多样性，可有效模拟实际应用场景中缺陷检测所面临的各类挑战。每个缺陷类别均包含300张图像，所有图像均为灰度图，分辨率为200×200像素。

搜集汇总

背景与挑战

背景概述

NEU-CLS数据集由东北大学表面检测实验室公开发布，专门用于带钢表面缺陷分类。该数据集包含六种常见带钢表面缺陷类型（Crazing、Scratch、Pitted Surface、Patch、Rolled-in Scale和Inclusion），每种缺陷有300张200×200像素的灰度图像，形态多样，具有高复杂性和多样性，能有效模拟实际应用场景中的缺陷检测挑战。