Anomaly Detection on MVTec anomaly detection (test)
99.3AUC (object)DMAD
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DMAD2023.03 | 99.3 | 99.9 | 99.5 | — | |
| PatchCore2023.03 | 99.2 | 99 | 99.1 | — | |
| RD2023.03 | 98 | 99.5 | 98.5 | — | |
| DRAEM2023.03 | 97.4 | 99.1 | 98 | — | |
| CutPaste2023.03 | 95.5 | 97.5 | 96.1 | — | |
| PaDiM2023.03 | 93.8 | 98.8 | 95.5 | — | |
| PSVDD2023.03 | 90.8 | 94.5 | 92.1 | — | |
| RotNet (MLP head) + KDEBackbone=ResNet-18, Input Resolution=256x256, Detector Type=KDE, Projection Head=MLP2020.11 | 89 | 81 | 86.3 | — | |
| DAAD2023.03 | 88.8 | 91 | 89.5 | — | |
| DistAug ContrastiveBackbone=ResNet-18, Input Resolution=256x256, Data Augmentation=DistAug2020.11 | 88.6 | 82.5 | 86.5 | — | |
| RotNet + KDEBackbone=ResNet-18, Input Resolution=256x256, Detector Type=KDE2020.11 | 85.9 | 75.5 | 83.5 | — | |
| CAVGA RaBackbone=ResNet-18, Input Resolution=256x2562020.11 | 83.8 | 78.2 | 81.9 | — | |
| Vanilla ContrastiveBackbone=ResNet-18, Input Resolution=256x256, Data Augmentation=None2020.11 | 83.8 | 73 | 80.2 | — | |
| RotNetBackbone=ResNet-18, Input Resolution=256x256, Detector Type=Rotation Classifier2020.11 | 77.9 | 73.2 | 71 | — | |
| GN2023.03 | 75.5 | 77.5 | 76.2 | — | |
| ARNetData=full data, ImageNet Pretrain=false, Backbone=UNet2022.07 | — | — | — | 83.9 | |
| CflowADData=full data, ImageNet Pretrain=true, Backbone=WRN502022.07 | — | — | — | 98.3 | |
| CutPasteData=full data, ImageNet Pretrain=true, Backbone=Res182022.07 | — | — | — | 95.2 | |
| FYDData=full data, ImageNet Pretrain=true, Backbone=Res182022.07 | — | — | — | 97.3 | |
| GANomalyData=full data, ImageNet Pretrain=false, Backbone=UNet2022.07 | — | — | — | 80.5 | |
| MKDData=full data, ImageNet Pretrain=true, Backbone=Res182022.07 | — | — | — | 87.7 | |
| PaDiMData=full data, ImageNet Pretrain=true, Backbone=WRN502022.07 | — | — | — | 97.9 | |
| PatchCoreData=full data, ImageNet Pretrain=true, Backbone=WRN502022.07 | — | — | — | 99.1 | |
| RegAD (k=16)Data=16 images, ImageNet Pretrain=true, Backbone=Res182022.07 | — | — | — | 92.7 | |
| RegAD (k=32)Data=32 images, ImageNet Pretrain=true, Backbone=Res182022.07 | — | — | — | 94.6 | |
| RegAD (k=4)Data=4 images, ImageNet Pretrain=true, Backbone=Res182022.07 | — | — | — | 88.2 | |
| RegAD (k=8)Data=8 images, ImageNet Pretrain=true, Backbone=Res182022.07 | — | — | — | 91.2 |