Camera Pose Estimation on ScanNet 50 (500 frames)
0.08ATEpi^3
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| pi^3Dense settings=true, Full-attention baseline=true, Spd.=1.0×, Backbone=pi^32026.07 | 0.08 | 2.17 | 0.43 | 0.01 | — | |
| Fast pi^3Dense settings=true, Full-attention baseline=false, Spd.=2.2×, Backbone=pi^32026.07 | 0.09 | 2.35 | 0.41 | 0.02 | — | |
| A pi^3Dense settings=true, Full-attention baseline=false, Spd.=5.9×, Backbone=pi^32026.07 | 0.09 | 2.36 | 0.46 | 0.02 | — | |
| SAF3RDense settings=true, Full-attention baseline=false, Spd.=6.5×, Backbone=pi^32026.07 | 0.09 | 2.4 | 0.44 | 0.02 | — | |
| FastVGGT2026.06 | 0.093 | 4.016 | 0.642 | 0.031 | 41.4 | |
| RegimeVGGT2026.06 | 0.096 | 4.06 | 0.68 | 0.033 | 25.6 | |
| VGGT*Checkpointing=Activation checkpointing2026.06 | 0.105 | 4.645 | 0.693 | 0.032 | 130.2 | |
| S-VGGT2026.06 | 0.113 | 4.113 | 0.925 | 0.091 | 26.1 | |
| FastVGGTDense settings=true, Full-attention baseline=false, Spd.=3.2×, Backbone=VGGT2026.07 | 0.13 | 3.54 | 0.62 | 0.03 | — | |
| DA3Dense settings=true, Full-attention baseline=true, Spd.=1.0×, Backbone=DA32026.07 | 0.13 | 3.73 | 0.77 | 0.02 | — | |
| FastDA3Dense settings=true, Full-attention baseline=false, Spd.=3.1×, Backbone=DA32026.07 | 0.13 | 3.94 | 1 | 0.02 | — | |
| Sparse pi^3Dense settings=true, Full-attention baseline=false, Spd.=2.9×, Backbone=pi^32026.07 | 0.14 | 4.99 | 0.49 | 0.02 | — | |
| VGGTDense settings=true, Full-attention baseline=true, Spd.=1.0×, Backbone=VGGT2026.07 | 0.15 | 4.14 | 0.98 | 0.04 | — | |
| Co-MeDense settings=true, Full-attention baseline=false, Spd.=3.5×, Backbone=VGGT2026.07 | 0.15 | 4.02 | 1.19 | 0.05 | — | |
| SAF3RDense settings=true, Full-attention baseline=false, Spd.=4.6×, Backbone=VGGT2026.07 | 0.15 | 4.13 | 0.91 | 0.03 | — | |
| AVGGTDense settings=true, Full-attention baseline=false, Spd.=4.3×, Backbone=VGGT2026.07 | 0.16 | 4.31 | 1.11 | 0.04 | — | |
| SAF3RDense settings=true, Full-attention baseline=false, Spd.=3.8×, Backbone=DA32026.07 | 0.16 | 5.71 | 1.89 | 0.04 | — | |
| SparseVGGTDense settings=true, Full-attention baseline=false, Spd.=2.7×, Backbone=VGGT2026.07 | 0.18 | 5.49 | 1.3 | 0.04 | — | |
| ADA3Dense settings=true, Full-attention baseline=false, Spd.=3.8×, Backbone=DA32026.07 | 0.43 | 13.42 | 7.67 | 0.1 | — |