Monocular Depth Estimation on ScanNet
0.97AbsRelRVM-S
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RVM-SSize(M)=342025.12 | 0.97 | — | |
| SiamMAE-SSize(M)=272025.12 | 1 | — | |
| RVM-BSize(M)=1172025.12 | 1.08 | — | |
| 4DS-BSize(M)=91, config=dist from e2025.12 | 1.21 | — | |
| LeReSwith Ours=true2026.05 | 1.3 | 99.4 | |
| GeoWizardwith Ours=true2026.05 | 1.3 | 99.4 | |
| DPTwith Ours=true2026.05 | 1.4 | 99.4 | |
| Marigoldwith Ours=true2026.05 | 1.4 | 99.1 | |
| UniDepthV2with Ours=true2026.05 | 1.4 | 99.3 | |
| VideoMAE-BSize(M)=872025.12 | 1.5 | — | |
| MiDaSwith Ours=true2026.05 | 1.5 | 99.1 | |
| Lotuswith Ours=true2026.05 | 1.5 | 99.2 | |
| Metric3Dv2with Ours=true2026.05 | 1.5 | 99.2 | |
| DepthAnythingV2with Ours=true2026.05 | 1.6 | 99.1 | |
| DepthMasterwith Ours=true2026.05 | 1.6 | 99.2 | |
| 4DS-BSize(M)=912025.12 | 1.65 | — | |
| Depth Prowith Ours=true2026.05 | 2 | 98.6 | |
| 4DS-SSize(M)=242025.12 | 2.05 | — | |
| DepthLabSynthetic Training Samples=74k2024.12 | 2.3 | 98.5 | |
| Modality ForcingMethod Type=I2D from Joint Model2026.06 | 2.32 | 98.9 | |
| PPDType=Generative, Evaluation Protocol=Zero-shot2026.01 | 3.5 | 98.1 | |
| MoGe 2Type=Discriminative, Evaluation Protocol=Zero-shot2026.01 | 3.8 | 97.1 | |
| Depth ProType=Discriminative, Evaluation Protocol=Zero-shot2026.01 | 3.9 | 97.8 | |
| PPDMethod Type=Generative depth estimation2026.06 | 4.04 | 97.8 | |
| DepthAnythingV2Real Training Samples=62M, Synthetic Training Samples=595K2024.12 | 4.1 | 98.2 | |
| DepthAnything v2Zero-shot=true2025.02 | 4.2 | 97.8 | |
| Depth Anything v2Type=Discriminative, Evaluation Protocol=Zero-shot2026.01 | 4.2 | 97.6 | |
| DepthAnythingReal Training Samples=63.5M2024.12 | 4.2 | 98.1 | |
| Depth Prowith Ours=false2026.05 | 4.2 | 98 | |
| Depth AnythingTraining samples=63.5M, Feed-forward=true, Diffusion=false, Zero-shot=true2024.07 | 4.3 | 98.1 | |
| BetterDepth-2KTraining samples=2K, Feed-forward=true, Diffusion=true, Zero-shot=true2024.07 | 4.3 | 98 | |
| BetterDepthTraining samples=74K, Feed-forward=true, Diffusion=true, Zero-shot=true2024.07 | 4.3 | 98.1 | |
| Cross-Context DistillationZero-shot=true, Teacher Model=DepthAnythingv2-Large2025.02 | 4.3 | 98 | |
| DepthAnythingV2with Ours=false2026.05 | 4.3 | 98.1 | |
| BetterDepth-400Training samples=400, Feed-forward=true, Diffusion=true, Zero-shot=true2024.07 | 4.6 | 97.8 | |
| Metric3Dv2with Ours=false2026.05 | 4.7 | 99 | |
| Cross-Context DistillationZero-shot=true, Teacher Model=MiDaS v3.12025.02 | 4.9 | 98 | |
| HD-VGGT2026.03 | 4.9 | 96.1 | |
| LotusMethod Type=Generative depth estimation2026.06 | 5.01 | 96.5 | |
| WorldMirror2026.03 | 5.2 | 95.7 | |
| pi^32026.03 | 5.4 | 95.6 | |
| GeoWizardMethod Type=Generative depth estimation2026.06 | 5.49 | 95.7 | |
| DepthMasterwith Ours=false2026.05 | 5.5 | 96.7 | |
| GenPerceptType=Generative, Evaluation Protocol=Zero-shot2026.01 | 5.6 | 96.5 | |
| VGGT2026.03 | 5.6 | 95.1 | |
| JointDiTMethod Type=I2D from Joint Model2026.06 | 5.76 | 96.3 | |
| Marigold v1.1Prior=SD - 2.3B, training samples=74K2025.12 | 5.8 | 96.3 | |
| Marigold E2EPrior=SD - 2.3B, training samples=74K2025.12 | 5.8 | 96.2 | |
| UniDepthV2with Ours=false2026.05 | 5.8 | 97.5 | |
| MarigoldMethod Type=Generative depth estimation2026.06 | 5.89 | 95.2 | |
| LotusType=Generative, Evaluation Protocol=Zero-shot2026.01 | 5.9 | 95.7 | |
| Lotuswith Ours=false2026.05 | 6 | 96.3 | |
| GeoWizardTraining samples=280K, Feed-forward=false, Diffusion=true, Zero-shot=true2024.07 | 6.1 | 95.3 | |
| GeoWizardType=Generative, Evaluation Protocol=Zero-shot2026.01 | 6.1 | 95.3 | |
| GeoWizardSynthetic Training Samples=278K2024.12 | 6.1 | 95.4 | |
| GeoWizardwith Ours=false2026.05 | 6.1 | 95.3 | |
| GenPerceptZero-shot=true2025.02 | 6.3 | 96 | |
| DepthFMType=Generative, Evaluation Protocol=Zero-shot2026.01 | 6.3 | 95.4 | |
| MarigoldSynthetic Training Samples=74k2024.12 | 6.3 | 95.4 | |
| MarigoldTraining samples=74K, Feed-forward=false, Diffusion=true, Zero-shot=true2024.07 | 6.4 | 95.1 | |
| MarigoldZero-shot=true2025.02 | 6.4 | 95.1 | |
| MarigoldType=Generative, Evaluation Protocol=Zero-shot2026.01 | 6.4 | 95.1 | |
| Marigoldwith Ours=false2026.05 | 6.4 | 95.1 | |
| Metric3DPrior=DINOv2 - 142M, training samples=16M2025.12 | 6.6 | 95 | |
| MarigoldPrior=SD - 2.3B, training samples=74K2025.12 | 6.6 | 95 | |
| OmnidataReal Training Samples=11.9M, Synthetic Training Samples=310K2024.12 | 7.2 | 94.1 | |
| OmnidataTraining samples=12.2M, Feed-forward=true, Diffusion=false, Zero-shot=true2024.07 | 7.5 | 93.6 | |
| OmnidataType=Discriminative, Evaluation Protocol=Zero-shot2026.01 | 7.5 | 93.6 | |
| OmnidataZero-shot=true2025.02 | 7.7 | 93.5 | |
| DPTZero-shot=true2025.02 | 7.8 | 93.8 | |
| VAR-DepthPrior=Switti - 100M, training samples=74K, config=optimized wk2025.12 | 7.9 | 93.4 | |
| Hierarchical Depth NormalizationTraining Data=~ 300K, Zero-shot transfer=true2022.10 | 8 | 93.9 | |
| HDNTraining samples=300K, Feed-forward=true, Diffusion=false, Zero-shot=true2024.07 | 8 | 93.9 | |
| HDNZero-shot=true2025.02 | 8 | 93.9 | |
| HDNType=Discriminative, Evaluation Protocol=Zero-shot2026.01 | 8 | 93.9 | |
| HDNPrior=ImageNet - 14M, training samples=300K2025.12 | 8 | 93.9 | |
| HDNReal Training Samples=300K2024.12 | 8 | 94.2 | |
| DUSt3R2026.03 | 8.1 | 90.9 | |
| DPTType=Discriminative, Evaluation Protocol=Zero-shot2026.01 | 8.2 | 93.4 | |
| DPTPrior=ImageNet - 14M, training samples=1.4M2025.12 | 8.2 | 93.4 | |
| DPTTraining samples=1.4M, Feed-forward=true, Diffusion=false, Zero-shot=true2024.07 | 8.4 | 93.2 | |
| DPTReal Training Samples=1.2M, Synthetic Training Samples=188K2024.12 | 8.4 | 93.2 | |
| DPTwith Ours=false2026.05 | 8.4 | 93.2 | |
| LeReSTraining samples=354K, Feed-forward=true, Diffusion=false, Zero-shot=true2024.07 | 9.1 | 91.7 | |
| LeReSType=Discriminative, Evaluation Protocol=Zero-shot2026.01 | 9.1 | 91.7 | |
| LeReSPrior=ImageNet - 1M, training samples=354K2025.12 | 9.1 | 91.7 | |
| LeReSwith Ours=false2026.05 | 9.1 | 91.7 | |
| UniConMethod Type=I2D from Joint Model2026.06 | 9.23 | 91 | |
| LeresTraining Data=~ 360K, Zero-shot transfer=true2022.10 | 9.5 | 91.2 | |
| LeReSZero-shot=true2025.02 | 9.5 | 91.2 | |
| LeReSReal Training Samples=300K, Synthetic Training Samples=54K2024.12 | 9.6 | 90.4 | |
| MiDaSTraining samples=2M, Feed-forward=true, Diffusion=false, Zero-shot=true2024.07 | 9.9 | 90.7 | |
| MiDaSwith Ours=false2026.05 | 9.9 | 90.7 | |
| DiverseDepthTraining Data=~ 300K, Zero-shot transfer=true2022.10 | 10.8 | 88.2 | |
| DiverseDepthZero-shot=true2025.02 | 10.8 | 88.2 | |
| DiverseDepthTraining samples=320K, Feed-forward=true, Diffusion=false, Zero-shot=true2024.07 | 10.9 | 88.2 | |
| DiverseDepthType=Discriminative, Evaluation Protocol=Zero-shot2026.01 | 10.9 | 88.2 | |
| DiverseDepthPrior=ImageNet - 1M, training samples=320K2025.12 | 10.9 | 88.2 | |
| MASt3R2026.03 | 11 | 86.5 | |
| MiDaSTraining Data=~ 2M, Zero-shot transfer=true2022.10 | 11.1 | 88.6 |