2026-10-14T13:02:15Z https://www.scidb.cn/oai

10.57760/sciencedb.38531 2026-06-11T10:44:49Z

2026-06-11 The associated data for the paper "Zero-Shot Anomaly Detection via Textual Decoupling" doi:10.57760/sciencedb.38531 en The paper conducted extensive experiments on twelve publicly available datasets to comprehensively evaluate the performance of the proposed method These datasets cover both industrial and medical fields In terms of industrial anomaly detection, six benchmark datasets including MVTec AD, VisA, MPDD, BTAD, DAGM, and DTD-synthetic were selected In the field of medicine, the paper uses the skin cancer detection dataset ISIC, colon polyp detection datasets CVC-ClinicDB and CVC-ColonDB, as well as brain tumor detection datasets Head_CT, BrainMRI, and Br35H In the evaluation of anomaly detection tasks, the paper adopts the Area Under the Receiver Operating Characteristic (AUROC) and Average Precision (AP) at the image level as the main evaluation indicators to comprehensively measure the performance of the model in distinguishing normal and abnormal images  For anomaly localization tasks, the paper uses pixel level AUROC and Area Under the PRO curve (AUPRO) as evaluation metrics  Zero-Shot Anomaly Detection; Textual Decoupling; AUROC zhao meng yang PUBLIC https://creativecommons.org/licenses/by-nc-nd/4.0/ dataset Science Data Bank