Image-level anomaly detection on MVTec AD pill
93.33Similar AUROCKNN
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| KNNBackbone=ResNet18, Seeds=252026.02 | 93.33 | — | — | — | 86.15 | |
| CoPBackbone=ResNet18, Seeds=252026.02 | 88.27 | — | — | — | 81.15 | |
| DICEBackbone=ResNet18, Seeds=252026.02 | 87.63 | — | — | — | 82.56 | |
| ResidualBackbone=ResNet18, Seeds=252026.02 | 86.09 | — | — | — | 68.97 | |
| ReActBackbone=ResNet18, Seeds=252026.02 | 83.72 | — | — | — | 80 | |
| KDEBackbone=ResNet18, Seeds=252026.02 | 81.35 | — | — | — | 73.59 | |
| WeiPerBackbone=ResNet18, Seeds=252026.02 | 80.77 | — | — | — | 80.29 | |
| GradNormBackbone=ResNet18, Seeds=252026.02 | 80.13 | — | — | — | 79.07 | |
| MCPBackbone=ResNet18, Seeds=252026.02 | 78.46 | — | — | — | 78.33 | |
| Mahal.Backbone=ResNet18, Seeds=252026.02 | 71.86 | — | — | — | 68.72 | |
| AnoDiff*Evaluator=UNet, Training Data=Generated defect images2025.03 | — | 98 | 100 | 97 | — | |
| AnoDiff‡Evaluator=UNet, Training Data=Generated defect images2025.03 | — | 97 | 99 | 95 | — | |
| DefectFillEvaluator=UNet, Training Data=Generated defect images2025.03 | — | 97 | 99 | 95 | — | |
| DFMGAN+Evaluator=UNet, Training Data=Generated defect images2025.03 | — | 92 | 97 | 92 | — |