Anomaly Detection on MVTec-AD Texture Categories (leave-one-domain-out)
0.991Carpet ScoreEISNet
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EISNetBackbone=ResNet-18, Learn-to-compare scheme=false2022.03 | 0.991 | 0.662 | 1 | 0.85 | 0.986 | 0.898 | |
| EISNet+Backbone=ResNet-18, Learn-to-compare scheme=true2022.03 | 0.982 | 0.728 | 1 | 0.858 | 0.979 | 0.909 | |
| Patch-based meta-learning modelBackbone=ResNet-18, Learn-to-compare scheme=true2022.03 | 0.943 | 0.73 | 1 | 0.956 | 0.962 | 0.918 | |
| Epi-FRCBackbone=ResNet-18, Learn-to-compare scheme=false2022.03 | 0.916 | 0.64 | 0.995 | 0.947 | 0.909 | 0.881 | |
| Epi-FRC+Backbone=ResNet-18, Learn-to-compare scheme=true2022.03 | 0.916 | 0.725 | 1 | 0.951 | 0.941 | 0.907 | |
| AGG+Backbone=ResNet-18, Learn-to-compare scheme=true2022.03 | 0.891 | 0.608 | 0.992 | 0.912 | 0.865 | 0.854 | |
| AGGBackbone=ResNet-18, Learn-to-compare scheme=false2022.03 | 0.875 | 0.628 | 0.981 | 0.886 | 0.852 | 0.845 |