Anomaly Detection on Road Anomaly
85.9APcDNP-SETR-L
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| cDNP-SETR-LOutlier Exposure=false2022.11 | 85.9 | — | — | 13.8 | |
| cDNP-Segmenter-BOutlier Exposure=false2022.11 | 85.6 | — | — | 9.8 | |
| RbAOutlier Exposure=false2022.11 | 78.5 | — | — | 11.8 | |
| Proposed single-reference anomaly detection methodBackbone=DINOv3 [16], Training Protocol=training-free, Reference Setup=single-reference, Resolution=fixed spatial resolution2026.05 | 70.83 | — | 92.83 | 39.82 | |
| M2F-EAMOutlier Exposure=false2022.11 | 66.7 | — | — | 13.4 | |
| PEBALOutlier Exposure=true2022.11 | 62.4 | — | — | 28.3 | |
| MOoSeOutlier Exposure=false2022.11 | 43.6 | — | — | 32.1 | |
| DMLOutlier Exposure=false2022.11 | 37 | — | — | 37 | |
| Ours2021.07 | 25.82 | 80.33 | 81.96 | 49.74 | |
| SynthCPofficial code with pre-trained backbone=true, additional learnable parameters=true2021.07 | 24.86 | 80.33 | 76.08 | 64.69 | |
| Max Logit2021.07 | 24.44 | 80.33 | 77.97 | 64.85 | |
| Entropy2021.07 | 22.38 | 80.33 | 75.12 | 68.15 | |
| MSP2021.07 | 20.59 | 80.33 | 73.76 | 68.44 |