Misclassification detection on ImageNet-1K (val)
45.69FPR@95% (MSP)sample difficulty-aware entropy regularization
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| sample difficulty-aware entropy regularization2023.04 | 45.69 | 46.72 | 86.53 | 85.23 | 94.76 | 94.31 | |
| Confidence Ranking Loss2023.04 | 46.03 | 48.01 | 86.11 | 84.33 | 94.41 | 93.89 | |
| Cross Entropy2023.04 | 46.93 | 48.94 | 86.03 | 85.02 | 94.22 | 93.75 | |
| Entropy Regularization2023.04 | 46.98 | 48.85 | 86.15 | 84.19 | 94.38 | 93.82 | |
| Logit L12023.04 | 47.19 | 48.58 | 86.19 | 84.58 | 94.39 | 93.82 | |
| Focal Loss2023.04 | 48.54 | 51.1 | 85.67 | 83.05 | 94.08 | 93.03 | |
| Label Smoothing2023.04 | 52.12 | 61.52 | 85.28 | 81.03 | 93.86 | 91.98 |