热轧带钢表面缺陷数据集-东北大学

东北大学热轧带钢表面缺陷数据集，该数据集是东北大学的宋克臣等几位老师收集，该数据包括三个部分： 数据集收集了夹杂、划痕、压入氧化皮、裂纹、麻点和斑块6种缺陷，每种缺陷300张，图像尺寸为200×200。数据集包括分类和目标检测两部分，不过目标检测的标注中有少量错误，需要注意。部分示例如下： 微小型的带钢缺陷数据，缺陷只有约6×6个像素大小 油污干扰的硅钢表面缺陷数据集

Northeastern University Hot-Rolled Strip Steel Surface Defect Dataset
This dataset was collected by Professor Kechong Song and other researchers from Northeastern University. The dataset consists of three parts: it contains 6 types of defects including inclusions, scratches, pressed oxide scales, cracks, pitting spots and plaques, with 300 samples per defect type, and each image has a resolution of 200×200 pixels. The dataset includes two task subsets: image classification and object detection. Note that there are a small number of annotation errors in the object detection subset. Partial examples are as follows: 1) Ultra-small strip steel defects, where the defect size is only approximately 6×6 pixels; 2) Silicon steel surface defect dataset with oil stain interference.

搜集汇总

背景与挑战

背景概述

东北大学热轧带钢表面缺陷数据集包含三种不同类型的缺陷数据，主要用于分类和目标检测研究。数据集由6种常见缺陷图像组成，每种300张，尺寸为200×200，但需注意标注中可能存在少量错误。