DAGM2007

这是一个用于在纹理表面上进行缺陷检测的合成数据集。它最初是为 2007 年 DAGM 研讨会（Deutsche Arbeitsgemeinschaft für Mustererkennung eV，国际模式识别协会的德国分会）的竞赛而创建的。比赛与 GNSS（欧洲神经网络学会德国分会）共同主办。比赛结束后，该数据集已在多个项目和研究论文中用作测试数据集。它可从海德堡大学网站（Heidelberg Collaboratory for Image Processing）公开获得。数据是人工生成的，但类似于现实世界的问题。十个数据集中的前六个，称为开发数据集，应该用于算法开发。其余四个数据集，称为竞争数据集，可用于评估性能。研究人员应考虑在开发完成之前不使用或分析竞赛数据集作为荣誉守则。

This is a synthetic dataset designed for defect detection on textured surfaces. It was originally developed for the 2007 DAGM Symposium (Deutsche Arbeitsgemeinschaft für Mustererkennung eV, the German chapter of the International Association for Pattern Recognition) competition, which was co-hosted with GNSS (German Chapter of the European Neural Network Society). Following the conclusion of the competition, this dataset has been employed as a test dataset across numerous projects and research papers. It is publicly accessible via the website of Heidelberg University's Heidelberg Collaboratory for Image Processing. The dataset was artificially generated, yet mirrors real-world defect detection scenarios. Among the ten total datasets, the first six, referred to as the development datasets, should be used for algorithm development. The remaining four datasets, termed the competition datasets, are intended for performance evaluation. Researchers are requested to follow the honor code by refraining from using or analyzing the competition datasets before completing their algorithm development.

搜集汇总

背景与挑战

背景概述

DAGM2007是一个用于纹理表面缺陷检测的合成数据集，最初为2007年DAGM研讨会竞赛创建，由欧洲神经网络协会德国分会发布。该数据集包含十个子集，其中前六个为开发数据集用于算法开发，后四个为竞争数据集用于性能评估，数据虽为人工生成但模拟现实世界问题，并遵循荣誉守则建议在开发完成前不使用竞争数据集。