Deep Metric Learning on DyML-Product (overall)
47.9ASIHAPPIER
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HAPPIERTraining Strategy=Hierarchical2022.07 | 47.9 | 38 | 63.7 | — | — | |
| ΣNSMTraining Strategy=Hierarchical2022.07 | 46.9 | 36.8 | 60.8 | — | — | |
| ΣTLSHTraining Strategy=Hierarchical2022.07 | 46.3 | 36.9 | 58.5 | — | — | |
| TLSHTraining Strategy=Fine-grained2022.07 | 46.1 | 36.32 | 59.6 | — | — | |
| NSMTraining Strategy=Fine-grained2022.07 | 46 | 35.6 | 57.4 | — | — | |
| Smooth-APTraining Strategy=Fine-grained2022.07 | 45.5 | 36.1 | 55 | — | — | |
| ROADMAPTraining Strategy=Fine-grained2022.07 | 44.6 | 34.6 | 62.5 | — | — | |
| CSLTraining Strategy=Hierarchical2022.07 | 40.7 | 31.1 | 52.7 | — | — | |
| CSLTraining Supervision=multi-scale2021.03 | 29 | 28.7 | 54.3 | 83.1 | 89.4 | |
| Softmax LossTraining Supervision=multi-scale2021.03 | 26.8 | 26.1 | 50.2 | 81.6 | 87.7 | |
| Cosface LossTraining Supervision=multi-scale2021.03 | 25.5 | 25 | 49.3 | 81.3 | 87.7 | |
| Circle LossTraining Supervision=multi-scale2021.03 | 15.8 | 15 | 26.7 | 61.5 | 70.3 | |
| N-Pair LossTraining Supervision=multi-scale2021.03 | 15.7 | 15.3 | 20.3 | 55.5 | 65.6 | |
| MS LossTraining Supervision=multi-scale2021.03 | 9.8 | 10 | 12.7 | 45.7 | 56.4 | |
| Triplet LossTraining Supervision=multi-scale2021.03 | 9.2 | 9.3 | 11.2 | 43.6 | 53.3 |