基于异常特征对抗学习的工业图像异常检测方法
|
|
王天飞,周文俊,项圣,贺宇航,彭博
|
Industrial image anomaly detection method based on adversarial learning of abnormal features
|
|
Tianfei WANG,Wenjun ZHOU,Sheng XIANG,Yuhang HE,Bo PENG
|
|
| 表 5 消融实验图像级/像素级AUROC结果 |
| Tab.5 Image- and piexl-level AUROC results of ablation experiment |
|
| 类别 | AUROC | | 设计1 | 设计2 | 设计3 | | 1）注：斜线前、后数据分别表示图像级以及像素级AUROC结果 | | bottle | 98.2/69.71） | 96.8/97.2 | 97.2/98.5 | | capsule | 77.5/63.7 | 94.9/91.0 | 97.0/97.3 | | grid | 75.6/66.3 | 100.0/99.4 | 100.0/99.6 | | leather | 83.1/57.7 | 100.0/97.4 | 100.0/99.9 | | pill | 89.0/65.7 | 96.2/96.4 | 98.2/97.5 | | tile | 98.5/78.7 | 99.8/99.3 | 100.0/98.6 | | transistor | 91.4/59.6 | 92.7/88.9 | 96.3/92.1 | | zipper | 94.3/64.2 | 100.0/98.4 | 100.0/99.1 | | cable | 54.0/69.4 | 88.3/93.4 | 86.2/96.7 | | carpet | 52.5/56.1 | 90.6/93.8 | 97.6/99.2 | | hazelnut | 94.3/64.2 | 99.9/99.6 | 100.0/98.8 | | metalnut | 89.5/86.5 | 99.0/99.1 | 98.9/97.8 | | screw | 84.5/54.6 | 97.3/99.5 | 96.6/98.9 | | toothbrush | 86.6/76.9 | 100.0/97.7 | 100.0/97.9 | | wood | 91.0/67.9 | 99.6/94.9 | 97.6/94.5 | | 平均 | 84.0/66.75 | 97.0/96.4 | 98.2/97.8 |
|
|
|