Point cloud reconstruction on ScanNet++
17.2AccuracyADA3
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ADA3Sparsity=10-128 images2026.07 | 17.2 | 19.3 | 0.182 | 46.7 | |
| Sparsepi^3Sparsity=10-128 images2026.07 | 16.6 | 16.4 | 0.165 | 23.1 | |
| Co-MeSparsity=10-128 images2026.07 | 9.4 | 11.4 | 0.104 | 48.6 | |
| SparseVGGTSparsity=10-128 images2026.07 | 7.8 | 12 | 0.099 | 44.7 | |
| AVGGTSparsity=10-128 images2026.07 | 6.7 | 10.8 | 0.088 | 57.1 | |
| SAF3RComparison Context=VGGT baseline, Sparsity=10-128 images2026.07 | 6.6 | 10.8 | 0.087 | 56.2 | |
| Fastpi^3Sparsity=10-128 images2026.07 | 6.3 | 10.5 | 0.084 | 51.9 | |
| SAF3RComparison Context=DA3 baseline, Sparsity=10-128 images2026.07 | 5.7 | 9 | 0.074 | 67.3 | |
| FastVGGTSparsity=10-128 images2026.07 | 5.5 | 10.9 | 0.082 | 56.8 | |
| Api^3Sparsity=10-128 images2026.07 | 5.4 | 9.7 | 0.075 | 56.9 | |
| SAF3RComparison Context=pi^3 baseline, Sparsity=10-128 images2026.07 | 5.1 | 9.5 | 0.073 | 59.8 | |
| pi^3Attention=Full, Sparsity=10-128 images2026.07 | 4.7 | 9.2 | 0.07 | 64.9 | |
| VGGTAttention=Full, Sparsity=10-128 images2026.07 | 4.5 | 9.2 | 0.069 | 69.2 | |
| DA3Attention=Full, Sparsity=10-128 images2026.07 | 3.5 | 8 | 0.057 | 79.6 | |
| FastDA3Sparsity=10-128 images2026.07 | 3.4 | 8.2 | 0.058 | 78.2 |