Outlier Detection on ImageNet-30 (10% Labeled + Unlabeled)
96.3AUROC (In-Distribution)OpenMatch
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| OpenMatchTraining Supervision=Semi-supervised, Labeled Data Ratio=10%2021.05 | 96.3 | 89.9 | 84.4 | 87.7 | 80.8 | 87.7 | 92.1 | 87.1 | |
| MTCTraining Supervision=Semi-supervised, Labeled Data Ratio=10%2021.05 | 93.8 | 78 | 59.5 | 72.2 | 76.4 | 80.9 | 78 | 74.2 | |
| SupervisedTraining Supervision=Fully supervised, Training Condition=Trained with fully labeled inliers2021.05 | 92.8 | 94.4 | 92.7 | 91.5 | 88.2 | 89.9 | 92.3 | 91.3 | |
| FixMatchTraining Supervision=Semi-supervised, Labeled Data Ratio=10%2021.05 | 88.6 | 85.7 | 83.1 | 81 | 81.9 | 83.1 | 86.4 | 83 | |
| Labeled OnlyTraining Supervision=Semi-supervised, Labeled Data Ratio=10%2021.05 | 80.3 | 85.9 | 75.4 | 77.9 | 69 | 78.7 | 84.8 | 78.6 |