MVTec LOCO AD

该数据集名为MVTec LOCO AD，是专为工业异常检测设计的数据集。它包含了1,772张训练图像、304张验证图像以及1,568张测试图像，涵盖了五个更为复杂的检测场景。与MVTec AD数据集不同的是，该数据集还包括了两个难度较高的异常类别：逻辑异常和结构异常。总的来说，该数据集在规模上拥有1,772张训练图像、304张验证图像以及1,568张测试图像，其任务专注于工业异常检测。

This dataset, named MVTec LOCO AD, is specifically designed for industrial anomaly detection. It comprises 1,772 training images, 304 validation images, and 1,568 test images, covering five relatively complex detection scenarios. Unlike the MVTec AD dataset, this collection also includes two high-difficulty anomaly categories: logical anomalies and structural anomalies. In summary, this dataset contains 1,772 training images, 304 validation images, and 1,568 test images, with its core task focused on industrial anomaly detection.

搜集汇总

背景与挑战

背景概述

MVTec LOCO AD数据集是一个用于无监督异常定位算法评估的工业检测数据集，包含3644张图像，涵盖结构性和逻辑性异常，并提供像素级地面实况数据。该数据集适用于非商业用途，并附带评估代码以支持研究。