OOD Detection on CIFAR-10 vs SVHN
99.9ROCAUCDeep Clustering for Multi-Class Anomaly Detection
Evaluation Results
| Method | Links | |
|---|---|---|
| Deep Clustering for Multi-Class Anomaly DetectionNetwork=ViT, Pre-trained=true2021.12 | 99.9 | |
| CSINetwork=ResNet-18, Pre-trained=false2021.12 | 99.8 | |
| Deep Clustering for Multi-Class Anomaly DetectionNetwork=ResNet-152, Pre-trained=true2021.12 | 98.8 | |
| MSCLNetwork=ResNet-152, Pre-trained=true2021.12 | 98.6 | |
| Deep Clustering for Multi-Class Anomaly DetectionNetwork=ResNet-18, Pre-trained=true2021.12 | 98.6 | |
| RotNetwork=ResNet-18, Pre-trained=false2021.12 | 97.8 | |
| GOADNetwork=ResNet-18, Pre-trained=false2021.12 | 96.3 | |
| DN2Network=ResNet-152, Pre-trained=true2021.12 | 96.2 | |
| Input ComplexityNetwork=Glow, Pre-trained=false2021.12 | 95 | |
| DN2Network=ResNet-18, Pre-trained=true2021.12 | 95 | |
| SCAN FeaturesNetwork=ResNet-18, Pre-trained=false2021.12 | 94.3 | |
| Likelihood RatioNetwork=PixelCNN++, Pre-trained=false2021.12 | 91.2 | |
| LikelihoodNetwork=Glow, Pre-trained=false2021.12 | 8.3 |