Point Cloud Reconstruction on HiRoom
25.7AccuracySAF3R
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SAF3RComparison Context=VGGT baseline, Sparsity=10-128 images2026.07 | 25.7 | 18.4 | 0.22 | 45.9 | |
| SparseVGGTSparsity=10-128 images2026.07 | 24.6 | 20.3 | 0.225 | 27.9 | |
| Sparsepi^3Sparsity=10-128 images2026.07 | 23.4 | 14.5 | 0.19 | 20.3 | |
| Co-MeSparsity=10-128 images2026.07 | 21 | 13.4 | 0.172 | 40.6 | |
| FastVGGTSparsity=10-128 images2026.07 | 18.1 | 16.1 | 0.171 | 42.9 | |
| ADA3Sparsity=10-128 images2026.07 | 16.3 | 9.7 | 0.13 | 63.1 | |
| AVGGTSparsity=10-128 images2026.07 | 15.6 | 12.3 | 0.14 | 46.7 | |
| Fastpi^3Sparsity=10-128 images2026.07 | 12.7 | 9.9 | 0.113 | 35.2 | |
| VGGTAttention=Full, Sparsity=10-128 images2026.07 | 10.7 | 9.2 | 0.1 | 55.4 | |
| Api^3Sparsity=10-128 images2026.07 | 9.9 | 6.3 | 0.081 | 56.1 | |
| SAF3RComparison Context=pi^3 baseline, Sparsity=10-128 images2026.07 | 9.2 | 7.4 | 0.083 | 50.1 | |
| FastDA3Sparsity=10-128 images2026.07 | 6 | 6 | 0.06 | 76.2 | |
| pi^3Attention=Full, Sparsity=10-128 images2026.07 | 5.3 | 4.3 | 0.048 | 73.7 | |
| SAF3RComparison Context=DA3 baseline, Sparsity=10-128 images2026.07 | 5.3 | 4.8 | 0.051 | 84.2 | |
| DA3Attention=Full, Sparsity=10-128 images2026.07 | 4.8 | 4.4 | 0.046 | 85.9 |