Cross-domain Calibration on miniImageNet to CUB (test)
0.9ECEProto Nets
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Proto Nets2022.08 | 0.9 | 2.5 | 60.4 | |
| TRIDENT2022.08 | 0.9 | 2 | 27.6 | |
| OVE(PL)Objective=Pseudo-Likelihood2022.08 | 2 | 3.2 | 55.6 | |
| LogSoftGP(PL)Objective=Pseudo-Likelihood2022.08 | 2.2 | 4.2 | 56.4 | |
| Matching Nets2022.08 | 3 | 7.9 | 63 | |
| BMAMLVariation=Base2022.08 | 4.8 | 7.7 | 61.9 | |
| OVE(ML)Objective=Marginal Likelihood2022.08 | 4.9 | 6.6 | 57.6 | |
| BMAML+ChaserVariation=Chaser2022.08 | 6.6 | 26 | 63.9 | |
| LogSoftGP(ML)Objective=Marginal Likelihood2022.08 | 22 | 51.3 | 70.9 | |
| Relation Net2022.08 | 23.4 | 55.4 | 73 | |
| DKT+CosKernel=Cosine2022.08 | 23.6 | 42.6 | 67 | |
| Feature Transfer2022.08 | 27.5 | 64.6 | 77.2 | |
| Baseline2022.08 | 31.5 | 53.7 | 71.6 |