Image Retrieval on Revisited Oxford (ROxf) Hard 1.0 (test)
60.2mAPDELG + RRT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DELG + RRTTraining set=GLDv2-clean, Net=ResNet-50, Number of local descriptors=500, Feature Type=Global features + Re-ranking2021.03 | 60.2 | 44.1 | |
| RRTDesc. version=R50-v2-clean, # local desc.=500, Desc. dim.=1282021.03 | 60.2 | — | |
| SuperGlueDesc. version=SuperPoint [17], # local desc.=1024, Desc. dim.=256, Global Retrieval Baseline=DELG R50-v2-clean2021.03 | 60 | — | |
| DELG + GVTraining set=GLDv2-clean, Net=ResNet-50, Number of local descriptors=1000, Feature Type=Global features + Re-ranking2021.03 | 57.9 | 43.6 | |
| HOW-ASMKTraining set=SfM-120k, Net=ResNet-50, Number of local descriptors=2000, Feature Type=Local feature aggregation2021.03 | 56.9 | 38.9 | |
| DELG + RRTTraining set=GLDv1&v2-clean, Net=ResNet-50, Number of local descriptors=500, Feature Type=Global features + Re-ranking2021.03 | 56.4 | 37 | |
| RRTDesc. version=R50-v1, # local desc.=500, Desc. dim.=1282021.03 | 56.4 | — | |
| HOW-ASMKTraining set=SfM-120k, Net=ResNet-50, Number of local descriptors=1000, Feature Type=Local feature aggregation2021.03 | 55.8 | 36.8 | |
| SuperGlueDesc. version=SuperPoint [17], # local desc.=1024, Desc. dim.=256, Global Retrieval Baseline=DELG R50-v12021.03 | 55.2 | — | |
| SuperGlueDesc. version=SuperPoint [17], # local desc.=500, Desc. dim.=256, Global Retrieval Baseline=DELG R50-v2-clean2021.03 | 54.6 | — | |
| DELG + GVTraining set=GLDv1, Net=ResNet-50, Number of local descriptors=1000, Feature Type=Global features + Re-ranking2021.03 | 54.2 | 36.8 | |
| DELG globalDesc. version=R50-v2-clean, # local desc.=02021.03 | 51 | — | |
| SuperGlueDesc. version=SuperPoint [17], # local desc.=500, Desc. dim.=256, Global Retrieval Baseline=DELG R50-v12021.03 | 49.7 | — | |
| DELGTraining set=GLDv1, Net=ResNet-50, Number of local descriptors=0, Feature Type=Global features2021.03 | 45.1 | 27.8 | |
| DELG globalDesc. version=R50-v1, # local desc.=02021.03 | 45.1 | — | |
| DELF-ASMKTraining set=Landmarks, Net=ResNet-50, Number of local descriptors=1000, Feature Type=Local feature aggregation2021.03 | 43.1 | 31.2 | |
| GeM-APTraining set=SfM-120k, Net=ResNet-101, Number of local descriptors=0, Feature Type=Global features2021.03 | 42.8 | 23.2 | |
| GeM + DSMTraining set=SfM-120k, Net=ResNet-101, Number of local descriptors=1000, Feature Type=Global features + Re-ranking2021.03 | 39.2 | 23.2 | |
| GeMTraining set=SfM-120k, Net=ResNet-101, Number of local descriptors=0, Feature Type=Global features2021.03 | 38.5 | 19.9 | |
| R-MACTraining set=Landmarks, Net=ResNet-101, Number of local descriptors=0, Feature Type=Global features2021.03 | 32.4 | 12.5 |