训练样本

[SorryMaker666]

您好，在您的论文中提到We employ one-third of the anomaly data with the lowest ID numbers as the training set, reserving the remaining two-thirds for testing，想请问您一下，这里的the anomaly data是使用了test文件夹和ground_truth文件夹下的数据吗，因为我看到train文件夹下的数据都是good无缺陷的数据

[engrmusawarali]

@SorryMaker666 MVTec AD provides the test set for defects, so in anomaly diffusion they split the test set into two parts.