基于多任务对比约束与可控缺陷生成的异常检测算法
刘宇轩,苏恩祈,井佩光,孙宏博,苏育挺

Anomaly detection algorithm based on multi-task contrastive constraints and controllable anomaly synthesis
Yuxuan LIU,Enqi SU,Peiguang JING,Hongbo SUN,Yuting SU
表 4 不同算法在BTAD数据集中的图像级和像素级异常检测结果
Tab.4 Image-level and pixel-level anomaly detection results of different algorithms on BTAD dataset
算法AUROCI/%AUROCP/%
Class 01Class 02Class 03平均值Class 01Class 02Class 03平均值
VT-ADL99.094.077.090.097.671.082.683.7
P-SVDD91.693.691.092.195.772.182.183.3
FastFlow99.482.491.190.997.193.698.396.3
PatchCore95.594.799.396.590.979.399.890.0
CFA98.185.599.094.295.996.098.696.8
AWSSA-UNet93.391.999.094.710082.099.893.9
本研究96.593.999.796.796.296.499.797.4