Image Retrieval on Stanford Online Products
88.4Recall@1GAPan
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GAPanArch=ViT2026.05 | 88.4 | 96.6 | 98.9 | 99.6 | — | |
| LaFGArch=ViT2026.05 | 87.1 | 95.8 | 98.5 | 99.5 | — | |
| SEEIArch=ViT2026.05 | 86.3 | 95 | 98.2 | — | — | |
| HIERArch=ViT2026.05 | 86.1 | 95 | 98 | — | — | |
| DDMLArch=ViT2026.05 | 86.1 | 95.1 | 98.2 | 99.5 | — | |
| HypViTArch=ViT2026.05 | 85.9 | 94.9 | 98.1 | 99.5 | — | |
| DPHMArch=ViT2026.05 | 84.8 | 94.5 | 98.1 | 99.4 | — | |
| VPTSP-GICLRArch=ViT2026.05 | 84.4 | 93.6 | 97.3 | — | — | |
| CGD (SG/GS)Backbone=SE-ResNet-50, Dim=1536, Input Size=2562019.03 | 84.2 | 93.9 | 97.4 | 99.2 | — | |
| DFML-PACArch=ViT2026.05 | 84.2 | 93.8 | — | — | — | |
| CGD (SG/GS)Backbone=ResNet-50, Dim=1536, Input Size=2562019.03 | 83.9 | 93.8 | 97.5 | 99.2 | — | |
| BFE+Backbone=ResNet-50, Dim=1536, Input Size=2562019.03 | 83 | 93.3 | 97.3 | 99.2 | — | |
| HISTArch=R502026.05 | 81.4 | 92 | 96.7 | — | — | |
| PNCA++Arch=R502026.05 | 81.4 | 92.4 | 96.9 | 99 | — | |
| CGD (SG/GS)Backbone=ResNet-50, Dim=128, Input Size=2242019.03 | 81 | 92.2 | 96.8 | 99.1 | — | |
| CGD (SG/-)Backbone=BN-Inception, Dim=512, Input Size=2242019.03 | 80.5 | 92.1 | 96.7 | 98.9 | — | |
| NIRArch=R502026.05 | 80.4 | 91.4 | — | — | — | |
| contrastive loss + HORDEBackbone=BN-Inception2019.08 | 80.1 | 91.3 | 96.2 | 98.7 | — | |
| HSEArch=R502026.05 | 80 | 91.4 | 96.3 | — | — | |
| CBMLArch=R502026.05 | 79.9 | 91.5 | 96.5 | 98.9 | — | |
| RLLEnsemble size=3, Category=Loss + Ensembles2019.12 | 79.8 | 91.3 | 96.3 | — | — | |
| DIMLArch=ViT2026.05 | 79.5 | — | — | — | — | |
| CGD (SG/GS)Backbone=ShuffleNet-v2, Dim=1536, Input Size=2242019.03 | 78.7 | 90.9 | 96.1 | 98.8 | — | |
| EPSHNEmbedding Dimension=5122019.04 | 78.3 | 90.7 | 96.3 | — | — | |
| Multi-similarity lossBackbone=BN-Inception2019.08 | 78.2 | 90.5 | 96 | 98.7 | — | |
| GPWCategory=Loss + Sampling2019.12 | 78.2 | 90.5 | 96 | — | — | |
| ABEEmbedding Dimension=5122019.04 | 76.3 | 88.4 | 94.8 | — | — | |
| ABE-8Backbone=GoogleNet, Dim=512, Input Size=2562019.03 | 76.3 | 88.4 | 94.8 | 98.2 | — | |
| ABEEnsemble size=8, Category=Loss + Ensembles2019.12 | 76.3 | 88.4 | 94.8 | — | — | |
| Group LossEnsemble size=5, Category=Loss + Ensembles2019.12 | 76.3 | 88.3 | 94.6 | — | — | |
| RLLEnsemble size=1, Category=Loss + Sampling2019.12 | 76.1 | 89.1 | 95.4 | — | — | |
| D and CEnsemble size=8, Category=Loss + Ensembles2019.12 | 75.9 | 88.4 | 94.9 | — | — | |
| Group LossEnsemble size=2, Category=Loss + Ensembles2019.12 | 75.9 | 88 | 94.5 | — | — | |
| CGD (SG/GS)Backbone=BN-Inception, Dim=64, Input Size=2242019.03 | 75.6 | 89 | 95.5 | 98.6 | — | |
| ABEEnsemble size=2, Category=Loss + Ensembles2019.12 | 75.4 | 88 | 94.7 | — | — | |
| RKDStrategy=Teacher-Student2019.12 | 75.1 | 88.3 | 95.2 | — | — | |
| HTLBackbone=GoogleNet2019.08 | 74.8 | 88.3 | 94.8 | 98.4 | — | |
| HTLEmbedding Dimension=5122019.04 | 74.8 | 88.3 | 94.8 | — | — | |
| HTLBackbone=BN-Inception, Dim=512, Input Size=2242019.03 | 74.8 | 88.3 | 94.8 | 98.4 | — | |
| Hier. tripletCategory=Loss + Sampling2019.12 | 74.8 | 88.3 | 94.8 | — | — | |
| A-BIEREnsemble size=6, Category=Loss + Ensembles2019.12 | 74.2 | 86.9 | 94 | — | — | |
| Margin2017.06 | 72.7 | 86.2 | 93.8 | 98 | 90.7 | |
| BIEREmbedding Dimension=5122019.04 | 72.7 | 86.5 | 94 | — | — | |
| MarginBackbone=ResNet-50, Dim=128, Input Size=2242019.03 | 72.7 | 86.2 | 93.8 | 98 | — | |
| Samp. Matt.Category=Loss + Sampling2019.12 | 72.7 | 86.2 | 93.8 | — | — | |
| BIEREnsemble size=6, Category=Loss + Ensembles2019.12 | 72.7 | 86.5 | 94 | — | — | |
| Binomial Deviance + HORDEBackbone=GoogleNet2019.08 | 72.6 | 85.9 | 93.7 | 97.9 | — | |
| Angular lossBackbone=GoogleNet2019.08 | 70.9 | 85 | 93.5 | 98 | — | |
| DVMLBackbone=GoogleNet2019.08 | 70.2 | 85.2 | 93.8 | — | — | |
| HDCEnsemble size=3, Category=Loss + Ensembles2019.12 | 70.1 | 84.9 | 93.2 | — | — | |
| DAMLRMMBackbone=GoogleNet2019.08 | 69.7 | 85.2 | 93.2 | — | — | |
| DAMLRRMCategory=Loss + Sampling2019.12 | 69.7 | 85.2 | 93.2 | — | — | |
| HDC2017.06 | 69.5 | 84.4 | 92.8 | 97.7 | — | |
| HDCEmbedding Dimension=3842019.04 | 69.5 | 84.4 | 92.8 | — | — | |
| DE-DSPCategory=Loss + Sampling2019.12 | 68.9 | 84 | 92.6 | — | — | |
| HDMLBackbone=GoogleNet2019.08 | 68.7 | 83.2 | 92.4 | — | — | |
| N-pairs2017.06 | 67.7 | 83.8 | 93 | 97.8 | 88.1 | |
| Binomial Deviance (Ours)Backbone=GoogleNet2019.08 | 67.4 | 81.7 | 90.2 | 95.4 | — | |
| StructClustering2017.06 | 67 | 83.7 | 93.2 | — | 89.5 | |
| FacilityBackbone=BN-Inception, Dim=64, Input Size=2242019.03 | 67 | 83.7 | 93.2 | — | — | |
| Triplet Semi-hard2017.06 | 66.7 | 82.4 | 91.9 | — | 89.5 | |
| Binomial Deviance2017.06 | 65.5 | 82.3 | 92.3 | 97.6 | — | |
| Histogram2017.06 | 63.9 | 81.7 | 92.2 | 97.7 | — | |
| LiftedStruct2017.06 | 62.5 | 80.8 | 91.9 | — | 88.7 |