Point-Level Anomaly Segmentation on STU (val)
99.53AUROCNDP-EE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NDP-EEAuxiliary OOD Data=✓, Architecture=Mask4Former2026.04 | 99.53 | 1.43 | 74.24 | |
| NDP-EnergyAuxiliary OOD Data=✓, Architecture=Mask4Former2026.04 | 99.37 | 2.35 | 66.54 | |
| NDP-EntropyAuxiliary OOD Data=✓, Architecture=Mask4Former2026.04 | 98.85 | 6.89 | 27.9 | |
| LIDO2026.04 | 95.05 | 34.86 | 27.53 | |
| Deep EnsembleAuxiliary OOD Data=✗, Architecture=Mask4Former2026.04 | 90.93 | 37.34 | 6.94 | |
| Mask4Former3D + Deep EnsembleBase Architecture=Mask4Former3D, Anomaly Detection Method=Deep Ensemble2026.04 | 90.93 | 37.34 | 6.94 | |
| Void ClassifierAuxiliary OOD Data=✓, Architecture=Mask4Former2026.04 | 89.77 | 79.5 | 2.62 | |
| Mask4Former3D + Void ClassifierBase Architecture=Mask4Former3D, Anomaly Detection Method=Void Classifier2026.04 | 89.77 | 79.5 | 2.62 | |
| MaxLogitAuxiliary OOD Data=✗, Architecture=Mask4Former2026.04 | 87.27 | 68.76 | 2.02 | |
| Mask4Former3D + Max LogitBase Architecture=Mask4Former3D, Anomaly Detection Method=Max Logit2026.04 | 87.27 | 68.76 | 2.02 | |
| RbAAuxiliary OOD Data=✗, Architecture=Mask4Former2026.04 | 73 | 100 | 1.64 | |
| Mask4Former3D + RbABase Architecture=Mask4Former3D, Anomaly Detection Method=RbA2026.04 | 73 | 100 | 1.64 | |
| MC DropoutAuxiliary OOD Data=✗, Architecture=Mask4Former2026.04 | 65.76 | 79.82 | 0.17 | |
| Mask4Former3D + MC DropoutBase Architecture=Mask4Former3D, Anomaly Detection Method=MC Dropout2026.04 | 65.76 | 79.82 | 0.17 |