Deep Metric Learning on DyML Animal (overall)
50.8ASIHAPPIER
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HAPPIERTraining Strategy=Hierarchical2022.07 | 50.8 | 43.8 | 68.9 | — | — | |
| ΣNSMTraining Strategy=Hierarchical2022.07 | 50.6 | 42.6 | 70 | — | — | |
| NSMTraining Strategy=Fine-grained2022.07 | 48.4 | 38.8 | 69.6 | — | — | |
| ΣTLSHTraining Strategy=Hierarchical2022.07 | 47.2 | 38.9 | 65.9 | — | — | |
| TLSHTraining Strategy=Fine-grained2022.07 | 46.3 | 37.5 | 66.3 | — | — | |
| CSLTraining Strategy=Hierarchical2022.07 | 46.3 | 40.8 | 60.9 | — | — | |
| Smooth-APTraining Strategy=Fine-grained2022.07 | 45.4 | 37.7 | 63.6 | — | — | |
| ROADMAPTraining Strategy=Fine-grained2022.07 | 42.6 | 34.4 | 62.8 | — | — | |
| N-Pair LossTraining Supervision=multi-scale2021.03 | 0.457 | 0.303 | 0.396 | 0.696 | 0.788 | |
| CSLTraining Supervision=multi-scale2021.03 | 0.452 | 0.31 | 0.523 | 0.817 | 0.883 | |
| Circle LossTraining Supervision=multi-scale2021.03 | 0.445 | 0.306 | 0.415 | 0.722 | 0.803 | |
| Cosface LossTraining Supervision=multi-scale2021.03 | 0.396 | 0.284 | 0.451 | 0.757 | 0.833 | |
| Softmax LossTraining Supervision=multi-scale2021.03 | 0.359 | 0.258 | 0.496 | 0.817 | 0.888 | |
| MS LossTraining Supervision=multi-scale2021.03 | 0.199 | 0.116 | 0.167 | 0.535 | 0.648 | |
| Triplet LossTraining Supervision=multi-scale2021.03 | 0.193 | 0.11 | 0.182 | 0.555 | 0.663 |