Point cloud reconstruction on 7-Scenes
2Inference Time (s)TurboVGGT
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| TurboVGGTStride=102026.05 | 2 | 1.6 | 0.9 | 2.5 | 1 | 65 | 73 | |
| SparseVGGTStride=102026.05 | 2.6 | 2 | 1 | 2.8 | 1 | 62.9 | 70.7 | |
| FastVGGTStride=102026.05 | 2.7 | 1.8 | 0.8 | 2.7 | 1 | 63.2 | 71.7 | |
| RegimeVGGTStride=102026.06 | 3 | 1.7 | 0.8 | 2.8 | 1.2 | 61.1 | 67.6 | |
| CUT3RStride=102026.06 | 3.2 | 4.1 | 2.1 | 2.9 | 1 | 64 | 70.7 | |
| FastVGGTStride=102026.06 | 3.5 | 1.7 | 0.8 | 2.8 | 1.2 | 61.1 | 67.7 | |
| S-VGGTStride=102026.06 | 4 | 2.5 | 1.2 | 2.5 | 1.1 | 61.9 | 68.8 | |
| Fast3RStride=102026.06 | 4.1 | 4 | 2.1 | 5.6 | 1.3 | 62.8 | 68.7 | |
| VGGTStride=102026.05 | 4.5 | 1.9 | 0.8 | 2.7 | 1 | 62.8 | 71 | |
| LiteVGGTInput Sequence Length=100, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 4.5 | — | — | — | — | — | — | |
| Fast3RStride=102026.05 | 5.5 | 4 | 2.1 | 5.6 | 1.3 | 63.9 | 65.7 | |
| Co-MeInput Sequence Length=100, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 5.5 | — | — | — | — | — | — | |
| Sparse-π³Input Sequence Length=100, Sparsification Rate (SR)=75%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 5.7 | — | — | — | — | — | — | |
| VGGT*Stride=102026.06 | 6.5 | 1.9 | 0.8 | 2.9 | 1.3 | 60.9 | 67.5 | |
| Sparse-π³Input Sequence Length=100, Sparsification Rate (SR)=50%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 6.8 | — | — | — | — | — | — | |
| GoToHuntInput Sequence Length=100, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 7.8 | — | — | — | — | — | — | |
| FastVGGTInput Sequence Length=100, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 9.4 | — | — | — | — | — | — | |
| TurboVGGTStride=32026.05 | 9.6 | 1.6 | 0.9 | 2.6 | 1 | 63.9 | 71.4 | |
| LiteVGGTInput Sequence Length=200, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 10.1 | — | — | — | — | — | — | |
| CUT3RStride=32026.06 | 10.9 | 17.9 | 12.1 | 9.7 | 4.3 | 57.7 | 61.1 | |
| VGGT (Base Model)Input Sequence Length=100, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 13.6 | — | — | — | — | — | — | |
| RegimeVGGTStride=32026.06 | 13.9 | 1.7 | 0.8 | 2.7 | 1.2 | 60.2 | 66.1 | |
| Sparse-π³Input Sequence Length=200, Sparsification Rate (SR)=75%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 14 | — | — | — | — | — | — | |
| FastVGGTStride=32026.05 | 14.2 | 1.8 | 0.8 | 2.6 | 1 | 62.7 | 70.7 | |
| GoToHuntInput Sequence Length=200, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 15.9 | — | — | — | — | — | — | |
| SparseVGGTStride=32026.05 | 16.2 | 2.1 | 1 | 2.9 | 1 | 62.1 | 69.3 | |
| Co-MeInput Sequence Length=200, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 16.4 | — | — | — | — | — | — | |
| S-VGGTStride=32026.06 | 17.1 | 2.1 | 1 | 2.1 | 0.9 | 60.2 | 66.3 | |
| LiteVGGTInput Sequence Length=300, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 17.8 | — | — | — | — | — | — | |
| FastVGGTStride=32026.06 | 18.1 | 1.7 | 0.8 | 2.7 | 1.2 | 60.3 | 66.2 | |
| Sparse-π³Input Sequence Length=200, Sparsification Rate (SR)=50%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 18.3 | — | — | — | — | — | — | |
| AVGGTStride=32026.05 | 20.6 | 5.4 | 3.1 | 13.1 | 7.8 | 52.8 | 54.1 | |
| FastVGGTInput Sequence Length=200, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 23.1 | — | — | — | — | — | — | |
| GoToHuntInput Sequence Length=300, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 23.8 | — | — | — | — | — | — | |
| Sparse-π³Input Sequence Length=300, Sparsification Rate (SR)=75%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 25.6 | — | — | — | — | — | — | |
| LiteVGGTInput Sequence Length=400, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 26 | — | — | — | — | — | — | |
| GoToHuntInput Sequence Length=400, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 31.7 | — | — | — | — | — | — | |
| Co-MeInput Sequence Length=300, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 32.2 | — | — | — | — | — | — | |
| Fast3RStride=32026.06 | 33 | 4.5 | 2.7 | 4.7 | 1 | 60.5 | 65.7 | |
| Sparse-π³Input Sequence Length=300, Sparsification Rate (SR)=50%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 35 | — | — | — | — | — | — | |
| LiteVGGTInput Sequence Length=500, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 36.5 | — | — | — | — | — | — | |
| VGGTStride=32026.05 | 38.1 | 1.9 | 0.9 | 2.7 | 1 | 62.2 | 69.9 | |
| Sparse-π³Input Sequence Length=400, Sparsification Rate (SR)=75%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 39.5 | — | — | — | — | — | — | |
| FastVGGTInput Sequence Length=300, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 40 | — | — | — | — | — | — | |
| GoToHuntInput Sequence Length=500, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 41.2 | — | — | — | — | — | — | |
| Fast3RStride=32026.05 | 43.7 | 4.5 | 2.7 | 4.7 | 1 | 61.6 | 62.7 | |
| VGGT (Base Model)Input Sequence Length=200, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 47.4 | — | — | — | — | — | — | |
| Co-MeInput Sequence Length=400, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 53.3 | — | — | — | — | — | — | |
| Sparse-π³Input Sequence Length=500, Sparsification Rate (SR)=75%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 55.4 | — | — | — | — | — | — | |
| VGGT*Stride=32026.06 | 56 | 1.8 | 0.8 | 2.8 | 1.2 | 60 | 65.8 | |
| Sparse-π³Input Sequence Length=400, Sparsification Rate (SR)=50%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 56.2 | — | — | — | — | — | — | |
| FastVGGTInput Sequence Length=400, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 59.5 | — | — | — | — | — | — | |
| Sparse-π³Input Sequence Length=500, Sparsification Rate (SR)=50%, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 80.3 | — | — | — | — | — | — | |
| Co-MeInput Sequence Length=500, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 84.2 | — | — | — | — | — | — | |
| FastVGGTInput Sequence Length=500, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 84.6 | — | — | — | — | — | — | |
| VGGT (Base Model)Input Sequence Length=300, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 101.3 | — | — | — | — | — | — | |
| VGGT (Base Model)Input Sequence Length=400, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 179.8 | — | — | — | — | — | — | |
| VGGT (Base Model)Input Sequence Length=500, Hardware (GPU)=NVIDIA L40S GPU, K=25, sigma=32026.05 | 288 | — | — | — | — | — | — | |
| FGQBitwidth (W/A)=4/82026.05 | — | 4.3 | 2.4 | 5.5 | 3.1 | 73.6 | 84.8 | |
| FGQBitwidth (W/A)=4/42026.05 | — | 4.8 | 3.1 | 5.9 | 3.6 | 72.3 | 83.2 | |
| FlatQuantBitwidth (W/A)=4/82026.05 | — | 4.5 | 2.5 | 5.5 | 3.1 | 73.2 | 84.5 | |
| FlatQuantBitwidth (W/A)=4/42026.05 | — | 5.6 | 3.6 | 7 | 4.7 | 71.7 | 82.4 | |
| FP16Bitwidth (W/A)=16/162026.05 | — | 4.4 | 2.4 | 5.6 | 3.3 | 73.3 | 84.6 | |
| QuantVGGTBitwidth (W/A)=4/82026.05 | — | 4.4 | 2.5 | 5.6 | 3.3 | 72.9 | 84 | |
| QuantVGGTBitwidth (W/A)=4/42026.05 | — | 5.3 | 3.2 | 8.5 | 5.9 | 71.9 | 82.8 | |
| QuaRotBitwidth (W/A)=4/82026.05 | — | 4.7 | 2.7 | 5.6 | 3.3 | 73.6 | 84.8 | |
| QuaRotBitwidth (W/A)=4/42026.05 | — | 10.7 | 7.9 | 16.5 | 12.8 | 66.3 | 74.9 | |
| RTNBitwidth (W/A)=4/82026.05 | — | 4.6 | 2.7 | 6 | 3.7 | 72.9 | 83.9 | |
| RTNBitwidth (W/A)=4/42026.05 | — | 14.6 | 10.3 | 13.4 | 8.6 | 60 | 65.5 |