Image Retrieval on Revisited Paris (RPar) (Hard)
88.4mAPSuperGlobal
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SuperGlobalBackbone=ResNet-101, Strategy=SuperGlobal feature retrieval and reranking, Reranking Depth=16002023.08 | 88.4 | — | — | |
| SuperGlobalBackbone=ResNet-101, Strategy=SuperGlobal feature retrieval and reranking, Reranking Depth=8002023.08 | 88.2 | — | — | |
| SuperGlobalBackbone=ResNet-101, Retrieval Category=SuperGlobal retrieval and reranking, Reranking Setting=Rerank top 4002023.08 | 86.7 | — | — | |
| SuperGlobalBackbone=ResNet-101, Strategy=SuperGlobal feature retrieval and reranking, Reranking Depth=4002023.08 | 86.7 | — | — | |
| SuperGlobalBackbone=ResNet-50, Strategy=SuperGlobal feature retrieval and reranking, Reranking Depth=16002023.08 | 86.4 | — | — | |
| SuperGlobalBackbone=ResNet-50, Strategy=SuperGlobal feature retrieval and reranking, Reranking Depth=8002023.08 | 86.2 | — | — | |
| SuperGlobalBackbone=ResNet-50, Retrieval Category=SuperGlobal retrieval and reranking, Reranking Setting=Rerank top 4002023.08 | 84.4 | — | — | |
| SuperGlobalBackbone=ResNet-50, Strategy=SuperGlobal feature retrieval and reranking, Reranking Depth=4002023.08 | 84.4 | — | — | |
| SuperGlobalBackbone=ResNet-101, Retrieval Category=Global feature retrieval, Reranking Setting=No reranking2023.08 | 83.5 | — | — | |
| SuperGlobalBackbone=ResNet-101, Strategy=Global feature retrieval2023.08 | 83.5 | — | — | |
| CVNetBackbone=ResNet-101, Retrieval Category=Global feature retrieval + Local feature reranking, Reranking Setting=Rerank top 4002023.08 | 81.1 | — | — | |
| CVNetBackbone=ResNet-101, Strategy=Global feature retrieval + Local feature reranking, Reranking Depth=4002023.08 | 81.1 | — | — | |
| DOLGBackbone=ResNet-101, Retrieval Category=Global feature retrieval, Reranking Setting=No reranking2023.08 | 80.3 | — | — | |
| SuperGlobalBackbone=ResNet-50, Retrieval Category=Global feature retrieval, Reranking Setting=No reranking2023.08 | 80.3 | — | — | |
| DOLGBackbone=ResNet-101, Strategy=Global feature retrieval2023.08 | 80.3 | — | — | |
| SuperGlobalBackbone=ResNet-50, Strategy=Global feature retrieval2023.08 | 80.3 | — | — | |
| CVNetBackbone=ResNet-50, Retrieval Category=Global feature retrieval + Local feature reranking, Reranking Setting=Rerank top 4002023.08 | 80.2 | — | — | |
| CVNetBackbone=ResNet-50, Strategy=Global feature retrieval + Local feature reranking, Reranking Depth=4002023.08 | 80.2 | — | — | |
| CVNetBackbone=ResNet-101, Retrieval Category=Global feature retrieval, Reranking Setting=No reranking2023.08 | 79.1 | — | — | |
| CVNetBackbone=ResNet-101, Strategy=Global feature retrieval2023.08 | 79.1 | — | — | |
| RRT# Reranked images=400, Desc. version=R50-v2-clean2021.03 | 78.1 | — | — | |
| alphaQE + RRT# Reranked images=200, Desc. version=R50-v2-clean2021.03 | 77.7 | — | — | |
| DOLGBackbone=ResNet-50, Retrieval Category=Global feature retrieval, Reranking Setting=No reranking2023.08 | 77.7 | — | — | |
| DOLGBackbone=ResNet-50, Strategy=Global feature retrieval2023.08 | 77.7 | — | — | |
| RRT# Reranked images=200, Desc. version=R50-v2-clean2021.03 | 77.1 | — | — | |
| RRT# local desc.=500, # Reranked images=200, Desc. version=R50-v2-clean2021.03 | 77.1 | — | — | |
| CVNetBackbone=ResNet-50, Retrieval Category=Global feature retrieval, Reranking Setting=No reranking2023.08 | 76.5 | — | — | |
| CVNetBackbone=ResNet-50, Strategy=Global feature retrieval2023.08 | 76.5 | — | — | |
| CiDeR-FTTraining Set=RGLDv2-clean, Fine-tuned=true2024.04 | 75.3 | 90 | — | |
| RRTDesc. version=R50-v2-clean, # local desc.=500, Desc. dim.=1282021.03 | 75.1 | — | — | |
| RRT# Reranked images=100, Desc. version=R50-v2-clean2021.03 | 75.1 | — | — | |
| RRT# local desc.=500, # Reranked images=100, Desc. version=R50-v2-clean2021.03 | 75.1 | — | — | |
| RRT# Reranked images=400, Desc. version=R50-v12021.03 | 74.1 | — | — | |
| alphaQEDesc. version=R50-v2-clean2021.03 | 73.2 | — | — | |
| alphaQE + RRT# Reranked images=200, Desc. version=R50-v12021.03 | 72.8 | — | — | |
| DELGBackbone=ResNet-101, Retrieval Category=Global feature retrieval + Local feature reranking, Reranking Setting=GV rerank top 1002023.08 | 72.8 | — | — | |
| DELGBackbone=ResNet-101, Strategy=Global feature retrieval + Local feature reranking, Reranking Depth=1002023.08 | 72.8 | — | — | |
| DELGBackbone=ResNet-101, Retrieval Category=Global feature retrieval, Reranking Setting=No reranking2023.08 | 72.4 | — | — | |
| DELGBackbone=ResNet-101, Strategy=Global feature retrieval2023.08 | 72.4 | — | — | |
| RRT# Reranked images=200, Desc. version=R50-v12021.03 | 71.6 | — | — | |
| RRT# local desc.=500, # Reranked images=200, Desc. version=R50-v12021.03 | 71.6 | — | — | |
| DELG globalDesc. version=R50-v2-clean, # local desc.=02021.03 | 71.5 | — | — | |
| DELG globalDesc. version=R50-v2-clean2021.03 | 71.5 | — | — | |
| DELG global# local desc.=0, # Reranked images=0, Desc. version=R50-v2-clean2021.03 | 71.5 | — | — | |
| DELGBackbone=ResNet-50, Retrieval Category=Global feature retrieval, Reranking Setting=No reranking2023.08 | 71.5 | — | — | |
| DELGBackbone=ResNet-50, Strategy=Global feature retrieval2023.08 | 71.5 | — | — | |
| GV# local desc.=1000, # Reranked images=100, Desc. version=R50-v2-clean2021.03 | 71 | — | — | |
| DELGBackbone=ResNet-50, Retrieval Category=Global feature retrieval + Local feature reranking, Reranking Setting=GV rerank top 1002023.08 | 71 | — | — | |
| DELGBackbone=ResNet-50, Strategy=Global feature retrieval + Local feature reranking, Reranking Depth=1002023.08 | 71 | — | — | |
| SuperGlueDesc. version=SuperPoint [17], # local desc.=1024, Desc. dim.=256, Global Retrieval Baseline=DELG R50-v2-clean2021.03 | 70.4 | — | — | |
| CiDeR-FTTrain Set=SfM-120k, Backbone=ResNet-101, Pooling=GeM, Loss=ArcFace, Fine-tuned=true, Dimension=20482024.04 | 68.9 | — | — | |
| CiDeR-FTTraining Set=SfM-120k, Fine-tuned=true2024.04 | 68.9 | 91.3 | — | |
| Yokoo et al.Training Set=RGLDv2-clean, Fine-tuned=false2024.04 | 68.7 | 86.4 | — | |
| RRTDesc. version=R50-v1, # local desc.=500, Desc. dim.=1282021.03 | 68.6 | — | — | |
| RRT# Reranked images=100, Desc. version=R50-v12021.03 | 68.6 | — | — | |
| RRT# local desc.=500, # Reranked images=100, Desc. version=R50-v12021.03 | 68.6 | — | — | |
| CiDeRTraining Set=RGLDv2-clean, Fine-tuned=false2024.04 | 68.5 | 89.1 | — | |
| SuperGlueDesc. version=SuperPoint [17], # local desc.=500, Desc. dim.=256, Global Retrieval Baseline=DELG R50-v2-clean2021.03 | 68.2 | — | — | |
| DOLGTraining Set=RGLDv2-clean, Fine-tuned=false2024.04 | 67.9 | 87.4 | — | |
| GV*# local desc.=1000, # Reranked images=200, Desc. version=R50-v2-clean2021.03 | 67.2 | — | — | |
| alphaQEDesc. version=R50-v12021.03 | 66.1 | — | — | |
| GLAMTraining Set=RGLDv2-clean, Fine-tuned=false2024.04 | 65.6 | 87.6 | — | |
| GV# local desc.=1000, # Reranked images=100, Desc. version=R50-v12021.03 | 64.9 | — | — | |
| SuperGlueDesc. version=SuperPoint [17], # local desc.=1024, Desc. dim.=256, Global Retrieval Baseline=DELG R50-v12021.03 | 64.6 | — | — | |
| DELG globalDesc. version=R50-v1, # local desc.=02021.03 | 63.4 | — | — | |
| DELG globalDesc. version=R50-v12021.03 | 63.4 | — | — | |
| DELG global# local desc.=0, # Reranked images=0, Desc. version=R50-v12021.03 | 63.4 | — | — | |
| GV*# local desc.=1000, # Reranked images=200, Desc. version=R50-v12021.03 | 63.2 | — | — | |
| SuperGlueDesc. version=SuperPoint [17], # local desc.=500, Desc. dim.=256, Global Retrieval Baseline=DELG R50-v12021.03 | 62.4 | — | — | |
| SOLARTraining Set=RGLDv2-clean, Fine-tuned=true2024.04 | 61.8 | 83.9 | — | |
| R-ASMK*Train Set=NC-clean, Backbone=ResNet-50, Pooling=Softmax, Local, Fine-tuned=true2024.04 | 57.7 | — | — | |
| CiDeRTrain Set=SfM-120k, Backbone=ResNet-101, Pooling=GeM, Loss=ArcFace, Fine-tuned=false, Dimension=20482024.04 | 57.5 | — | — | |
| CiDeRTraining Set=SfM-120k, Fine-tuned=false2024.04 | 57.5 | 87.1 | — | |
| AGEMTrain Set=SfM-120k, Backbone=ResNet-101, Pooling=GeM, Loss=Siamese, Fine-tuned=true, Dimension=20482024.04 | 57.3 | — | — | |
| AGEMTraining Set=SfM-120k, Fine-tuned=true2024.04 | 57.3 | — | — | |
| Radenovic et al.Train Set=SfM-120k, Backbone=ResNet-101, Pooling=GeM, Loss=Siamese, Fine-tuned=true, Dimension=20482024.04 | 56.3 | — | — | |
| Filip et al.Training Set=SfM-120k, Fine-tuned=true2024.04 | 56.3 | 89.1 | — | |
| DELF-ASMK*+SPTrain Set=SfM-120k, Backbone=ResNet-50, Pooling=Softmax, Fine-tuned=false2024.04 | 55.4 | — | — | |
| TokenTraining Set=RGLDv2-clean, Fine-tuned=false2024.04 | 54.8 | 81.3 | — | |
| GLAMTrain Set=SfM-120k, Backbone=ResNet-101, Pooling=GeM, Loss=ArcFace, Fine-tuned=true, Dimension=5122024.04 | 54.3 | — | — | |
| GLAMTraining Set=SfM-120k, Fine-tuned=false2024.04 | 54.3 | — | — | |
| R-ASMK*+SPTrain Set=NC-clean, Backbone=ResNet-50, Pooling=Softmax, Local, Fine-tuned=true2024.04 | 54 | — | — | |
| Supervisedep=100, Backbone=ResNet-502023.10 | 52.1 | — | — | |
| RN101+R-MACTrain Set=ImNet-1k, Method=PS2025.04 | 52.1 | — | — | |
| DINOv1-S/16Train Set=GLDv2, Method=PS2025.04 | 51.6 | — | — | |
| SOLARTrain Set=SfM-120k, Backbone=ResNet-101, Pooling=GeM, Loss=Triplet, SOS, Fine-tuned=true, Dimension=20482024.04 | 46.7 | — | — | |
| SOLARTraining Set=SfM-120k, Fine-tuned=true2024.04 | 46.7 | 83.6 | — | |
| DIRTrain Set=SfM-120k, Backbone=ResNet-101, Pooling=RMAC, Loss=Triplet, Fine-tuned=true, Dimension=20482024.04 | 42.4 | — | — | |
| DIRTraining Set=SfM-120k, Fine-tuned=false2024.04 | 42.4 | 83.6 | — | |
| DINO-vMFBackbone=ViT-Base/162024.05 | 39.5 | — | — | |
| iBOT-vMFBackbone=ViT-Base/162024.05 | 38.1 | — | — | |
| GeMTraining Set=SfM-120k, Fine-tuned=false2024.04 | 36.6 | 67.6 | — | |
| iBOTBackbone=ViT-Base/162024.05 | 36.6 | — | — | |
| DINOBackbone=ViT-Base/162024.05 | 35.8 | — | — | |
| HesAff-rSIFT-ASMK*+SPTrain Set=SfM-120k, Backbone=ResNet-50, Fine-tuned=false2024.04 | 35 | — | — | |
| DINOBackbone=ViT-Small/162024.05 | 34.5 | — | — | |
| DINOv1-S/16Train Set=ImNet-1k, Method=PS2025.04 | 34.4 | — | — | |
| DINO-vMFBackbone=ViT-Small/162024.05 | 33.4 | — | — | |
| MSNBackbone=ViT-Small/162024.05 | 33.3 | — | — | |
| MoCo-v3ep=1000, Backbone=ResNet-502023.10 | 31.3 | — | — |