Image Retrieval on SOP (test)
86Recall@1ROADMAP
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ROADMAPdim=384, Category=DeiT, Backbone=DeiT2021.10 | 86 | 94.4 | 97.6 | — | — | |
| IRTRdim=384, Category=DeiT, Backbone=DeiT2021.10 | 84.2 | 93.7 | 97.3 | — | — | |
| LOCORE-baseRe-ranking images count=1002025.03 | 83.8 | 92.9 | 96.9 | — | 71 | |
| LOCORE-smallRe-ranking images count=1002025.03 | 83.3 | 92.7 | 96.9 | — | 69.4 | |
| ROADMAPdim=512, Category=AP loss, Backbone=ResNet-502021.10 | 83.1 | 92.7 | 96.3 | — | — | |
| LOCORE-tinyRe-ranking images count=1002025.03 | 82.4 | 93.1 | 96.9 | — | 68 | |
| SCTEmbedding Dimension=512, Backbone=ResNet502020.07 | 81.9 | 92.6 | 96.8 | — | — | |
| Triplet SCTdim=512/64, Category=Metric learning, Backbone=ResNet-502021.10 | 81.9 | 92.6 | 96.8 | — | — | |
| RRTRe-ranking images count=1002025.03 | 81.9 | 92.4 | 96.9 | — | 67.2 | |
| ProxyNCA++dim=2048, Category=Classification, Backbone=ResNet-502021.10 | 81.4 | 92.4 | 96.9 | — | — | |
| SSR RerankRe-ranking images count=1002025.03 | 81.2 | 91.9 | 95.6 | — | 66.7 | |
| Cross-Entropydim=2048, Category=Classification, Backbone=ResNet-502021.10 | 81.1 | 91.7 | 96.3 | — | — | |
| aQERe-ranking images count=1002025.03 | 81.1 | 90.7 | 96.3 | — | 68.1 | |
| SHNEmbedding Dimension=512, Backbone=ResNet502020.07 | 81 | 92.3 | 96.8 | — | — | |
| Global descriptors2025.03 | 80.8 | 92.1 | 96.9 | — | 65.1 | |
| ProxyNCA++dim=512, Category=Classification, Backbone=ResNet-502021.10 | 80.7 | 92 | 96.7 | — | — | |
| XBMdim=128, Category=Metric learning, Backbone=ResNet-502021.10 | 80.6 | 91.6 | 96.2 | — | — | |
| SoftBin*dim=512, Category=AP loss, Backbone=ResNet-502021.10 | 80.6 | 91.3 | 96.1 | — | — | |
| HORDEdim=512, Category=Metric learning, Backbone=ResNet-502021.10 | 80.1 | 91.3 | 96.2 | — | — | |
| SmoothAPdim=512, Category=AP loss, Backbone=ResNet-502021.10 | 80.1 | 91.5 | 96.6 | — | — | |
| NSoftmaxdim=2048, Category=Classification, Backbone=ResNet-502021.10 | 79.5 | 91.5 | 96.7 | — | — | |
| SECdim=512, Category=Metric learning, Backbone=ResNet-502021.10 | 78.7 | 90.8 | 96.6 | — | — | |
| BlackBoxdim=512, Category=AP loss, Backbone=ResNet-502021.10 | 78.6 | 90.5 | 96 | — | — | |
| EasyPositiveEmbedding Dimension=512, Backbone=ResNet502020.07 | 78.3 | 90.7 | 96.3 | — | — | |
| MSEmbedding Dimension=512, Backbone=ResNet502020.07 | 78.2 | 90.5 | 96 | — | — | |
| MSdim=512, Category=Metric learning, Backbone=ResNet-502021.10 | 78.2 | 90.5 | 96 | — | — | |
| NSoftmaxdim=512, Category=Classification, Backbone=ResNet-502021.10 | 78.2 | 90.6 | 96.2 | — | — | |
| ProxyGMLdim=512, Category=Classification, Backbone=ResNet-502021.10 | 78 | 90.6 | 96.2 | — | — | |
| MICdim=512, Category=Metric learning, Backbone=ResNet-502021.10 | 77.2 | 89.4 | 95.6 | — | — | |
| AQERe-ranking images count=1002025.03 | 76.9 | 89.3 | 94.5 | — | 66.1 | |
| FastAPEmbedding Dimension=512, Backbone=ResNet502020.07 | 76.4 | 89 | 95.1 | — | — | |
| FastAPdim=512, Category=AP loss, Backbone=ResNet-502021.10 | 76.4 | 89 | 95.1 | — | — | |
| ABEEmbedding Dimension=512, Backbone=ResNet502020.07 | 76.3 | 88.4 | 94.8 | — | — | |
| ProxyNCAdim=512, Category=Classification, Backbone=ResNet-502021.10 | 73.7 | — | — | — | — | |
| BIEREmbedding Dimension=512, Backbone=ResNet502020.07 | 72.7 | 86.5 | 94 | — | — | |
| Triplet SHdim=512, Category=Metric learning, Backbone=ResNet-502021.10 | 72.7 | 86.2 | 93.8 | — | — | |
| DQERe-ranking images count=1002025.03 | 67.9 | 81.5 | 90.6 | — | 47.8 | |
| UDONPre-training=CLIP2024.06 | 67 | — | — | 82.1 | — | |
| UDONPre-training=ImageNet21k2024.06 | 64.2 | — | — | 80.2 | — | |
| LiftedStructdim=512, Category=Metric learning, Backbone=ResNet-502021.10 | 62.1 | 79.8 | 91.3 | — | — | |
| MLP baseline w/ Dyn SamplerPre-training=ImageNet21k2024.06 | 61.8 | — | — | 78.5 | — | |
| USC w/ Dyn SamplerPre-training=ImageNet21k2024.06 | 61.1 | — | — | 78.4 | — |