Cross-domain few-shot classification on mini-ImageNet → CUB (test)
0.007ECECDKT + Cosine (ML)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CDKT + Cosine (ML)shot=5, way=5, Loss=ML2023.10 | 0.007 | 0.02 | |
| ProtoNetshot=5, way=52023.10 | 0.009 | 0.025 | |
| CDKT + Cosine (PL)shot=5, way=5, Loss=PL2023.10 | 0.01 | 0.029 | |
| OVE PG GP + Cosine (PL)shot=5, way=5, Loss=PL2023.10 | 0.02 | 0.032 | |
| LS (Gibbs) + Cosine (PL)shot=5, way=5, Loss=PL2023.10 | 0.022 | 0.042 | |
| MatchingNetshot=5, way=52023.10 | 0.03 | 0.079 | |
| Bayesian MAMLshot=5, way=52023.10 | 0.048 | 0.077 | |
| OVE PG GP + Cosine (ML)shot=5, way=5, Loss=ML2023.10 | 0.049 | 0.066 | |
| Bayesian MAML (Chaser)shot=5, way=52023.10 | 0.066 | 0.26 | |
| LS (Gibbs) + Cosine (ML)shot=5, way=5, Loss=ML2023.10 | 0.22 | 0.513 | |
| RelationNetshot=5, way=52023.10 | 0.234 | 0.554 | |
| DKT + Cosineshot=5, way=52023.10 | 0.236 | 0.426 | |
| Feature Transfershot=5, way=52023.10 | 0.275 | 0.646 | |
| Baseline++shot=5, way=52023.10 | 0.315 | 0.537 |