Image-goal Navigation on Real-world Complex environment
85SR (%)MINav (Extended, 2h)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MINav (Extended, 2h)Visual backbones=DINOv3-ViT-L, Policy Params=2 M, Training Cost (TFLOPs-hours)=1502026.03 | 85 | 21 | 66 | |
| MINav (Extended, 2h)Visual backbones=DINOv3-ViT-S, Policy Params=2 M, Training Cost (TFLOPs-hours)=1502026.03 | 58 | 27 | 48 | |
| MINav (Minimal, 1h)Visual backbones=DINOv3-ViT-L, Policy Params=2 M, Training Cost (TFLOPs-hours)=1502026.03 | 38 | 31 | 30 | |
| GNMVisual backbones=MobileNetV2, Policy Params=9 M, Training Cost (TFLOPs-hours)=> 50002026.03 | 25 | 36 | 22 | |
| MINav (Minimal, 1h)Visual backbones=DINOv3-ViT-S, Policy Params=2 M, Training Cost (TFLOPs-hours)=1502026.03 | 25 | 33 | 21 | |
| NoMadVisual backbones=EfficientNet, Policy Params=19 M, Training Cost (TFLOPs-hours)=> 300002026.03 | 18 | 37 | 16 | |
| ViNTVisual backbones=EfficientNet, Policy Params=30 M, Training Cost (TFLOPs-hours)=300002026.03 | 8 | 38 | 7 | |
| RandomVisual backbones=N/A, Policy Params=N/A, Training Cost (TFLOPs-hours)=N/A2026.03 | 0 | 40 | 0 |