Multi-view Depth Estimation on ScanNet (test)
0.059Abs RelMVS2D (Ours-robust)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MVS2D (Ours-robust)Param (M)=24.4, Hardware=Nvidia V100 GPU, Variant=Robust2021.04 | 0.059 | 17.5 | 10.1 | 7.1 | — | |
| MVS2D (Ours)Param (M)=13.0, Hardware=Nvidia V100 GPU2021.04 | 0.059 | 42.9 | 29.1 | 21.8 | — | |
| RIAV-MVSGRU iterations (T)=24, Frame dimension=320 x 256, Time(fps)=3.77, Mem. (MB)=4297, Param. (M)=27.62022.05 | 0.0734 | — | — | — | — | |
| RIAV-MVSGRU iterations (T)=12, Frame dimension=320 x 256, Time(fps)=5.91, Mem. (MB)=4297, Param. (M)=27.62022.05 | 0.0752 | — | — | — | — | |
| RIAV-MVSGRU iterations (T)=8, Frame dimension=320 x 256, Time(fps)=6.98, Mem. (MB)=4297, Param. (M)=27.62022.05 | 0.076 | — | — | — | — | |
| ESTDFrame dimension=320 x 256, Time(fps)=14.08, Mem. (MB)=1799, Param. (M)=36.22022.05 | 0.0812 | — | — | — | — | |
| NASParam (M)=18.0, Hardware=Nvidia V100 GPU2021.04 | 0.086 | 0.9 | 0.6 | 0.4 | — | |
| FastMVSParam (M)=0.4, Hardware=Nvidia V100 GPU2021.04 | 0.089 | 9 | 6 | 4.3 | — | |
| MVSNetParam (M)=1.1, Hardware=Nvidia V100 GPU2021.04 | 0.094 | 4.1 | 2.4 | 1.6 | — | |
| DPSNetParam (M)=4.2, Hardware=Nvidia V100 GPU2021.04 | 0.094 | 1.1 | 0.7 | 0.5 | — | |
| IterMVSFrame dimension=320 x 256, Time(fps)=22.61, Mem. (MB)=2171, Param. (M)=0.342022.05 | 0.0991 | — | — | — | — | |
| BtsParam (M)=46.8, Hardware=Nvidia V100 GPU2021.04 | 0.117 | 17 | — | — | — | |
| PatchmatchNetParam (M)=0.2, Hardware=Nvidia V100 GPU2021.04 | 0.133 | 21.8 | 11.6 | 8.5 | — | |
| MVS2D (Ours-mono)Param (M)=12.3, Hardware=Nvidia V100 GPU, Variant=Monocular2021.04 | 0.145 | 94.7 | — | — | — | |
| DUSt3REncoder=ViT-Large2025.03 | 3.01 | — | — | — | 72.54 | |
| DUSt3R-512Resolution=512, In-domain training=true, Inference Time (s)=0.192025.03 | 3.1 | — | — | — | 71.8 | |
| MUSt3R-224Resolution=224, In-domain training=true, Inference Time (s)=0.322025.03 | 3.3 | — | — | — | 67.1 | |
| DUSt3R-224Resolution=224, In-domain training=true, Inference Time (s)=0.12025.03 | 4 | — | — | — | 59.8 | |
| MASt3REncoder=ViT-Large2025.03 | 4.17 | — | — | — | 65.22 | |
| Spann3RResolution=224, In-domain training=true, Inference Time (s)=0.292025.03 | 4.2 | — | — | — | 58.2 | |
| DUNEEncoder=ViT-Base2025.03 | 4.24 | — | — | — | 59.68 | |
| DeepV2DEncoder=Hourglass2025.03 | 4.4 | — | — | — | 54.8 | |
| MUSt3R-512Resolution=512, In-domain training=true, Inference Time (s)=0.192025.03 | 4.5 | — | — | — | 56.7 |