Anomaly Detection on VAD low-shot supervised 100 bad (train)
92.7Cl. AUROCAll AD + SegAD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| All AD + SegADSupervision Paradigm=SegAD (ours), Backbone AD model=PatchCore, FastFlow, RD4AD, and EfficientAD2024.05 | 92.7 | 36.8 | |
| PatchCore + SegADSupervision Paradigm=SegAD (ours), Backbone AD model=PatchCore2024.05 | 90.8 | 52.7 | |
| EfficientAD + SegADSupervision Paradigm=SegAD (ours), Backbone AD model=EfficientAD2024.05 | 90.6 | 40.1 | |
| RD4AD + SegADSupervision Paradigm=SegAD (ours), Backbone AD model=RD4AD2024.05 | 88.6 | 51.1 | |
| FastFlow + SegADSupervision Paradigm=SegAD (ours), Backbone AD model=FastFlow2024.05 | 84.3 | 64.5 | |
| DRASupervision Paradigm=Supervised Anomaly Detection2024.05 | 78.9 | 74.6 | |
| WRNSupervision Paradigm=Supervised Classifier, Backbone=Wide ResNet502024.05 | 78.7 | 77.9 | |
| DevNetSupervision Paradigm=Supervised Anomaly Detection2024.05 | 73.1 | 88.3 |