Pixel-level Anomaly Detection on DAGM
97.3AUROCCDO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CDOLearning Type=full-shot unsupervised2024.07 | 97.3 | 58.3 | — | |
| PatchCoreLearning Type=full-shot unsupervised2024.07 | 97.2 | 59.4 | — | |
| RD4ADCategory=Texture, |C|=10, Evaluation Protocol=Full-shot2023.10 | 96.8 | — | 91.9 | |
| PatchCoreCategory=Texture, |C|=10, Evaluation Protocol=Full-shot2023.10 | 95.9 | — | 87.9 | |
| AnomalyCLIPCategory=Texture, |C|=10, Evaluation Protocol=Zero-shot2023.10 | 95.6 | — | 91 | |
| AdaCLIPLearning Type=Zero-Shot (ZSAD)2024.07 | 91.5 | 57.5 | — | |
| AdaCLIPsupervised_training=true2024.07 | 91.5 | 57.5 | — | |
| DINOV2supervised_training=false2024.07 | 90.9 | 52 | — | |
| APRIL-GANsupervised_training=true2024.07 | 90.3 | 57.9 | — | |
| SAMsupervised_training=false2024.07 | 88.6 | 40.7 | — | |
| WinCLIPsupervised_training=false2024.07 | 81.3 | 13.9 | — | |
| SAAsupervised_training=false2024.07 | 62.7 | 32.6 | — |