Image Clustering on Stanford Online Products (test)
89.4NMIDAML
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DAMLbase_loss=N-pair2019.03 | 89.4 | 32.4 | |
| HDMLbase_loss=N-pair2019.03 | 89.3 | 32.2 | |
| N-pair2019.03 | 87.9 | 27.1 | |
| Angular2019.03 | 87.8 | 26.5 | |
| Lifted2019.03 | 87.2 | 25.3 | |
| HDMLbase_loss=Triplet2019.03 | 87.2 | 22.5 | |
| DAMLbase_loss=Triplet2019.03 | 87.1 | 22.3 | |
| Triplet hardmining=semi-hard negative2019.03 | 86.7 | 22.1 | |
| Triplet2019.03 | 86.3 | 20.2 | |
| DDML2019.03 | 83.4 | 10.7 | |
| Contrastive2019.03 | 82.4 | 10.1 |