Visual Place Recognition on Pitts250K P2E
1.2Query Time (ms)SwinT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SwinTBackbone=Swin-T, #Params. (M)=28.292025.06 | 1.2 | 10.1 | 26.3 | 36 | |
| ConvNeXtSBackbone=ConvNeXt-S, #Params. (M)=50.222025.06 | 1.2 | 14.2 | 28.8 | 39.6 | |
| Berton et al. (ViT-B)Backbone=ViT-B, #Params. (M)=86.862025.06 | 1.2 | 8 | 23 | 33 | |
| HypeVPR-BBackbone=Swin-T, #Params. (M)=28.292025.06 | 3.6 | 29.4 | 51.4 | 60.6 | |
| HypeVPR-BBackbone=ConvNeXt-S, #Params. (M)=50.222025.06 | 3.6 | 34.3 | 59 | 71.1 | |
| HypeVPR-OBackbone=ResNet-50, #Params. (M)=27.70, trained_on_extra_large_dataset=true2025.06 | 4 | 66.5 | 82.1 | 86.3 | |
| HypeVPR-LBackbone=Swin-T, #Params. (M)=28.292025.06 | 14 | 32.5 | 57.2 | 67.6 | |
| HypeVPR-LBackbone=ConvNeXt-S, #Params. (M)=50.222025.06 | 14 | 43.4 | 64.3 | 73.4 | |
| PanoVPRBackbone=Swin-T, #Params. (M)=28.29, sliding_window_crops=82025.06 | 17 | 22 | 42.2 | 51.8 | |
| PanoVPRBackbone=ConvNeXt-S, #Params. (M)=50.22, sliding_window_crops=82025.06 | 17 | 30.9 | 53.9 | 64.3 | |
| HypeVPR-BBackbone=ResNet-50, #Params. (M)=27.70, trained_on_extra_large_dataset=true2025.06 | 29.6 | 79.6 | 88.6 | 90.6 | |
| NetVLADBackbone=ResNet-50, #Params. (M)=25.302025.06 | 32.8 | 4 | 12.4 | 20 | |
| PanoVPRBackbone=Swin-T, #Params. (M)=28.29, sliding_window_crops=162025.06 | 48.6 | 33.6 | 56.7 | 66.4 | |
| PanoVPRBackbone=ConvNeXt-S, #Params. (M)=50.22, sliding_window_crops=162025.06 | 48.6 | 40.3 | 63 | 72.1 | |
| Orhan et al.Backbone=ResNet-101, #Params. (M)=136.62, trained_on_extra_large_dataset=true2025.06 | 1,555.2 | 47 | 66.4 | 73.6 |