Visual Place Recognition on Tokyo24/7
100Recall@1Pair-VPR-p
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Pair-VPR-pEnc. Size=ViT-G2024.10 | 100 | 100 | 100 | |
| EffoVPR-Rtwo-stage=true2024.05 | 98.7 | 98.7 | 98.7 | |
| DC-VLAQVenue & Year=-2026.01 | 98.7 | 99.7 | 99.7 | |
| EffoVPRTwo-stage reranking=true2025.09 | 98.7 | 98.7 | 98.7 | |
| QAABackbones=DINOv2-B, Cd=81922025.07 | 98.4 | — | — | |
| BoQBackbones=DINOv2-B, Cd=122882025.07 | 98.1 | — | — | |
| Pair-VPR-sEnc. Size=ViT-B2024.10 | 98.1 | 98.4 | 98.7 | |
| BoQVenue & Year=CVPR’242026.01 | 98.1 | 98.1 | 98.7 | |
| BoQRes=322, Two-stage=false2026.04 | 98.1 | 98.1 | 98.7 | |
| FoL++Res=322, Two-stage=true2026.04 | 98.1 | 99.1 | 99.1 | |
| QAABackbones=DINOv2-B, Cd=40962025.07 | 97.8 | — | — | |
| SAGE2025.09 | 97.5 | 99.1 | 99.4 | |
| QAABackbones=DINOv2-B, Cd=10242025.07 | 97.1 | — | — | |
| EDTFormerVenue & Year=TCSVT’252026.01 | 97.1 | 98.1 | 98.4 | |
| EDTFormer2025.09 | 97.1 | 98.1 | 98.4 | |
| EDTFormerRes=322, Two-stage=false2026.04 | 97.1 | 98.1 | 98.4 | |
| SAGERes=322, Two-stage=false2026.04 | 97.1 | 98.4 | 99.1 | |
| FoLRes=322, Two-stage=true2026.04 | 97.1 | 97.8 | 98.7 | |
| SALAD CMBackbones=DINOv2-B, Cd=84482025.07 | 96.8 | — | — | |
| SALAD-CMRes=322, Two-stage=false2026.04 | 96.8 | 97.5 | 97.8 | |
| FoL++ (global)Res=322, Two-stage=false2026.04 | 96.8 | 98.4 | 98.4 | |
| QAABackbones=DINOv2-B, Cd=20482025.07 | 96.5 | — | — | |
| MegaLocRes=322, Two-stage=false2026.04 | 96.5 | 98.1 | 99.7 | |
| FoL (global)Res=322, Two-stage=false2026.04 | 96.5 | 98.1 | 98.4 | |
| FoL-globalVenue & Year=AAAI’252026.01 | 96.2 | 98.7 | 98.7 | |
| VLAD-BuFF2025.09 | 96.2 | 98.7 | 99.4 | |
| ImAgeRes=322, Two-stage=false2026.04 | 96.2 | 98.1 | 98.4 | |
| DELGRe-ranking=true, Training Dataset=MSLS2023.03 | 95.9 | 96.8 | 97.1 | |
| SuperVLADRes=322, Two-stage=false2026.04 | 95.6 | 97.8 | 98.1 | |
| SALADEnc. Size=ViT-B2024.10 | 95.2 | 97.1 | 98.1 | |
| SALADtwo-stage=false2024.05 | 94.6 | 97.5 | 97.8 | |
| SALADVenue & Year=CVPR’242026.01 | 94.6 | 97.5 | 97.8 | |
| SALAD-CMVenue & Year=ECCV’242026.01 | 94.6 | 97.5 | 97.8 | |
| SALADRes=322, Two-stage=false2026.04 | 94.6 | 97.5 | 97.8 | |
| SelaVPRtwo-stage=true2024.05 | 94 | 96.8 | 97.5 | |
| SelaVPREnc. Size=ViT-L2024.10 | 94 | 96.8 | 97.5 | |
| SelaVPRVenue & Year=ICLR’242026.01 | 94 | 96.8 | 97.5 | |
| SelaVPRRes=322, Two-stage=true2026.04 | 94 | 96.8 | 97.5 | |
| EigenPlaceBackbones=ResNet-50, Cd=20482025.07 | 93 | — | — | |
| EigenPlacesBackbone=ResNet-50, Descriptor Dimension=20482023.08 | 93 | — | — | |
| EigenPlacesFeature Dimension=20482024.02 | 93 | 96.2 | 97.5 | |
| CricaVPRFeature Dimension=40962024.02 | 93 | 97.5 | 98.1 | |
| EigenPlacestwo-stage=false2024.05 | 93 | 96.2 | 97.5 | |
| CricaVPRtwo-stage=false2024.05 | 93 | 97.5 | 98.1 | |
| EigenPlacesVenue & Year=ICCV’232026.01 | 93 | 96.2 | 97.5 | |
| CricaVPRVenue & Year=CVPR’242026.01 | 93 | 97.5 | 98.1 | |
| EigenPlacesRes=322, Two-stage=false2026.04 | 93 | 96.2 | 97.5 | |
| AsymVPRQuery Network=EfficientViT-B2, Gallery Model Training Configuration=BoQ (DINOv2-B), Retrieval Protocol=Asymmetric2025.12 | 92.7 | 96.5 | — | |
| EigenPlacesEnc. Size=ResNet-502024.10 | 92.4 | 96.2 | 97.1 | |
| BoQtwo-stage=false2024.05 | 91.1 | 95.9 | 96.5 | |
| EffoVPR-ZSZero-Shot=true, re-ranking=top-100 global candidates2024.05 | 90.8 | — | — | |
| BoQEnc. Size=ResNet-502024.10 | 90.5 | 95.2 | 96.5 | |
| EigenPlacesBackbone=ResNet-50, Descriptor Dimension=5122023.08 | 89.8 | — | — | |
| CosPlaceBackbone=ResNet-50, Descriptor Dimension=5122023.08 | 89.5 | — | — | |
| DINOv2-RIALearning Protocol=training-free (with pre-trained backbone)2026.01 | 89.2 | 92.3 | — | |
| AsymVPRQuery Network=MobileViTv2, Gallery Model Training Configuration=BoQ (DINOv2-B), Retrieval Protocol=Asymmetric2025.12 | 89.2 | 94.3 | — | |
| R2 FormerDescriptor dimension=256, Train set=MSLS2024.06 | 88.6 | 91.4 | — | |
| ProGeo (Transformer)Descriptor dimension=512, Train set=SF-XL2024.06 | 88.6 | 93.3 | — | |
| R2Formertwo-stage=true2024.05 | 88.6 | 91.4 | 91.7 | |
| R2FormerEnc. Size=ViT-S2024.10 | 88.6 | 91.4 | 91.7 | |
| SuperGluereranking=100 candidates from NetVLAD2021.03 | 88.2 | 90.2 | 90.2 | |
| SP-SuperGlueRe-ranking=true, Training Dataset=MSLS2023.03 | 88.2 | 90.2 | 90.2 | |
| CosPlaceBackbones=ResNet-50, Cd=20482025.07 | 87.3 | — | — | |
| CosPlaceBackbone=ResNet-50, Descriptor Dimension=20482023.08 | 87.3 | — | — | |
| CosPlaceEnc. Size=ResNet-502024.10 | 87.3 | 94 | 95.6 | |
| ProGeo (CNN)Descriptor dimension=512, Train set=SF-XL2024.06 | 87 | 92.7 | — | |
| MixVPRtwo-stage=false2024.05 | 86.7 | 92.1 | 94 | |
| AsymVPRQuery Network=EfficientViT-B2, Gallery Model Training Configuration=SALAD (DINOv2-B), Retrieval Protocol=Asymmetric2025.12 | 86.7 | 94 | — | |
| PaVPR baselineEnc. Size=ViT-B2024.10 | 86.4 | 94.6 | 95.6 | |
| Patch-NetVLADspatial_matching=Multi-RANSAC2021.03 | 86 | 88.6 | 90.5 | |
| Patch NetVLADPCA=Yes, Dimension=4096, Re-ranking=true, Training Dataset=MSLS2023.03 | 86 | 88.6 | 90.5 | |
| Patch-NetVLADFeature Dimension=N/A2024.02 | 86 | 88.6 | 90.5 | |
| Patch-NetVLADtwo-stage=true2024.05 | 86 | 88.6 | 90.5 | |
| Patch-NetVLADVenue & Year=CVPR’212026.01 | 86 | 88.6 | 90.5 | |
| Patch-NetVLADRes=322, Two-stage=true2026.04 | 86 | 88.6 | 90.5 | |
| DINOv2-VLADLearning Protocol=Unsupervised Learning (with pre-trained backbone)2026.01 | 85.7 | 89.8 | — | |
| MixVPRBackbones=ResNet-50, Cd=40962025.07 | 85.1 | — | — | |
| MixVPRBackbone=ResNet-50, Descriptor Dimension=40962023.08 | 85.1 | — | — | |
| MixVPRFeature Dimension=40962024.02 | 85.1 | 91.7 | 94.3 | |
| MixVPREnc. Size=ResNet-502024.10 | 85.1 | 91.7 | 94.3 | |
| MixVPRLearning Protocol=Supervised Learning2026.01 | 85.1 | 91.7 | — | |
| MixVPRVenue & Year=WACV’232026.01 | 85.1 | 91.7 | 94.3 | |
| MixVPRRes=322, Two-stage=false2026.04 | 85.1 | 91.7 | 94.3 | |
| DINOv2-GeMLearning Protocol=training-free (with pre-trained backbone)2026.01 | 83.1 | 94.2 | — | |
| CosPlaceDesc. dim.=512, Train set=SF-XL, Backbone=VGG-162022.04 | 82.8 | 90 | — | |
| CosPlaceDescriptor dimension=512, Train set=SF-XL2024.06 | 82.8 | 90 | — | |
| EigenPlacesBackbone=VGG-16, Descriptor Dimension=5122023.08 | 82.2 | — | — | |
| CleanBackbone=ResNet18, Descriptor Bit Length=642026.06 | 82.2 | — | — | |
| CosPlaceBackbone=VGG-16, Descriptor Dimension=5122023.08 | 81.9 | — | — | |
| CosPlaceFeature Dimension=5122024.02 | 81.9 | 90.2 | 92.7 | |
| CosPlacetwo-stage=false2024.05 | 81.9 | 90.2 | 92.7 | |
| CosPlaceLearning Protocol=Supervised Learning2026.01 | 81.9 | 90.2 | — | |
| CosPlaceVenue & Year=CVPR’222026.01 | 81.9 | 90.2 | 92.7 | |
| CosPlaceRes=322, Two-stage=false2026.04 | 81.9 | 90.2 | 92.7 | |
| SelaVPRRes=322, Two-stage=false2026.04 | 81.9 | 94.9 | 96.5 | |
| Conv-APDescriptor dimension=512, Train set=GSV-Cities2024.06 | 81.3 | — | — | |
| CosPlaceBackbone=ResNet-50, Descriptor Dimension=1282023.08 | 81 | — | — | |
| SFRSFeature Dimension=40962024.02 | 81 | 88.3 | 92.4 | |
| SFRStwo-stage=false2024.05 | 81 | 88.3 | 92.4 | |
| SFRSVenue & Year=ECCV’202026.01 | 81 | 88.3 | 92.4 |