Monocular Depth Estimation on KITTI (AbsRel, delta1)
2.2AbsRelDepthAnythingV2
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DepthAnythingV2with Ours=true2026.05 | 2.2 | 98.7 | |
| Metric3Dv2with Ours=true2026.05 | 2.7 | 98 | |
| DPTwith Ours=true2026.05 | 3.2 | 97.6 | |
| Lotuswith Ours=true2026.05 | 3.2 | 97.5 | |
| UniDepthV2with Ours=true2026.05 | 3.2 | 97.7 | |
| Depth Prowith Ours=true2026.05 | 3.4 | 97.5 | |
| LeReSwith Ours=true2026.05 | 3.5 | 97.4 | |
| MoGe-2Method Type=Discriminative depth estimation2026.06 | 3.75 | 98.1 | |
| UniDepth V2Method Type=Discriminative depth estimation2026.06 | 3.85 | 98.1 | |
| Marigoldwith Ours=true2026.05 | 4.1 | 96.6 | |
| DepthMasterwith Ours=true2026.05 | 4.3 | 96.5 | |
| GeoWizardwith Ours=true2026.05 | 4.7 | 96.2 | |
| MoGe2025.07 | 4.9 | 97.9 | |
| MoGeversion=v12025.07 | 5 | 97.6 | |
| Metric3Dv2with Ours=false2026.05 | 5.1 | 97.6 | |
| Depth ProMethod Type=Discriminative depth estimation2026.06 | 5.12 | 96.8 | |
| Modality ForcingMethod Type=I2D from Joint Model2026.06 | 5.37 | 96.6 | |
| ViGeoAlignment=affine-invariant, Temporal Context=None2026.05 | 5.4 | 97.1 | |
| Pi3Alignment=affine-invariant2026.05 | 5.5 | 97.7 | |
| ZoeDepthMethod Type=Discriminative depth estimation2026.06 | 5.59 | 95.1 | |
| MASt3RMethod Type=Discriminative depth estimation2026.06 | 5.79 | 95.1 | |
| G2TAM2026.07 | 5.9 | 97.3 | |
| π32025.07 | 6 | 97.1 | |
| Pi32026.07 | 6 | 97.1 | |
| Omnizero-shot=true2026.04 | 6.21 | 96.92 | |
| MoGeversion=v22025.07 | 6.3 | 97.3 | |
| PPDMethod Type=Generative depth estimation2026.06 | 6.57 | 95.1 | |
| DA-v2Method Type=Discriminative depth estimation2026.06 | 6.77 | 94.3 | |
| MiDaSwith Ours=true2026.05 | 6.9 | 92.9 | |
| DepthLabSynthetic Training Samples=74k2024.12 | 7.2 | 95.3 | |
| DA3Alignment=affine-invariant2026.05 | 7.3 | 95.4 | |
| DepthAnything v2Prior=DINOv2 - 142M, training samples=63M2025.12 | 7.5 | 94.8 | |
| DepthAnythingV2Real Training Samples=62M, Synthetic Training Samples=595K2024.12 | 7.5 | 94.8 | |
| DepthAnythingReal Training Samples=63.5M2024.12 | 7.6 | 94.7 | |
| UniDepthV2with Ours=false2026.05 | 7.6 | 95.2 | |
| MASt3R2025.07 | 7.7 | 94.8 | |
| Depth Prowith Ours=false2026.05 | 7.7 | 94.9 | |
| VGGTAlignment=affine-invariant2026.05 | 7.7 | 92.9 | |
| MASt3R2026.07 | 7.7 | 94.8 | |
| DepthAnythingV2with Ours=false2026.05 | 8 | 94.6 | |
| VGGTzero-shot=true2026.04 | 8.03 | 94.29 | |
| VGGT2025.07 | 8.2 | 94.7 | |
| DepthMasterwith Ours=false2026.05 | 8.2 | 93.7 | |
| VGGT2026.07 | 8.2 | 94.7 | |
| DA3 giantzero-shot=true2026.04 | 8.24 | 93.96 | |
| DepthFMPrior=SD - 2.3B, training samples=63K2025.12 | 8.3 | 93.4 | |
| DepthFMSynthetic Training Samples=63K2024.12 | 8.4 | 93.2 | |
| FlashDepthAlignment=affine-invariant2026.05 | 8.4 | 93.9 | |
| LotusMethod Type=Generative depth estimation2026.06 | 8.89 | 90.1 | |
| MarigoldMethod Type=Generative depth estimation2026.06 | 9.05 | 90.3 | |
| Lotuswith Ours=false2026.05 | 9.3 | 92.8 | |
| Marigold E2EPrior=SD - 2.3B, training samples=74K2025.12 | 9.6 | 91.9 | |
| GeoWizardSynthetic Training Samples=278K2024.12 | 9.6 | 92.3 | |
| CUT3R2025.07 | 9.7 | 91.4 | |
| GeoWizardwith Ours=false2026.05 | 9.7 | 92.1 | |
| CUT3R2026.07 | 9.7 | 91.4 | |
| MonST3R2025.07 | 9.8 | 89.5 | |
| MarigoldSynthetic Training Samples=74k2024.12 | 9.8 | 91.7 | |
| MonST3R2026.07 | 9.8 | 89.5 | |
| Marigoldwith Ours=false2026.05 | 9.9 | 91.6 | |
| DPTPrior=ImageNet - 14M, training samples=1.4M2025.12 | 10 | 90.3 | |
| Metric3DPrior=DINOv2 - 142M, training samples=16M2025.12 | 10.1 | 90.9 | |
| MarigoldPrior=SD - 2.3B, training samples=74K2025.12 | 10.1 | 90.9 | |
| DPTReal Training Samples=1.2M, Synthetic Training Samples=188K2024.12 | 10.2 | 89.9 | |
| VAR-DepthPrior=Switti - 100M, training samples=74K, config=optimized wk2025.12 | 10.4 | 90.1 | |
| Marigold v1.1Prior=SD - 2.3B, training samples=74K2025.12 | 10.5 | 90.2 | |
| DUSt3R2025.07 | 10.9 | 87.3 | |
| JointDiTMethod Type=I2D from Joint Model2026.06 | 10.9 | 88.4 | |
| DUSt3R2026.07 | 10.9 | 87.3 | |
| GeoWizardMethod Type=Generative depth estimation2026.06 | 10.97 | 86.6 | |
| Marigoldzero-shot=true2026.04 | 11.08 | 87.87 | |
| DPTwith Ours=false2026.05 | 11.1 | 88.1 | |
| HDNReal Training Samples=300K2024.12 | 11.2 | 87.2 | |
| HDNPrior=ImageNet - 14M, training samples=300K2025.12 | 11.5 | 86.7 | |
| Fast3R2025.07 | 12 | 86.1 | |
| Fast3R2026.07 | 12 | 86.1 | |
| Cut3rzero-shot=true2026.04 | 12.53 | 86.42 | |
| JointNetMethod Type=I2D from Joint Model2026.06 | 13.74 | 81.3 | |
| OmnidataReal Training Samples=11.9M, Synthetic Training Samples=310K2024.12 | 14.7 | 83.7 | |
| LeReSPrior=ImageNet - 1M, training samples=354K2025.12 | 14.9 | 78.4 | |
| LeReSReal Training Samples=300K, Synthetic Training Samples=54K2024.12 | 14.9 | 78.5 | |
| LeReSwith Ours=false2026.05 | 14.9 | 78.4 | |
| MiDaSwith Ours=false2026.05 | 18.3 | 71.1 | |
| UniConMethod Type=I2D from Joint Model2026.06 | 18.67 | 73 | |
| DiverseDepthReal Training Samples=320K2024.12 | 18.8 | 70.2 | |
| DiverseDepthPrior=ImageNet - 1M, training samples=320K2025.12 | 19 | 70.4 | |
| DepthCrafterAlignment=affine-invariant2026.05 | 19.2 | 67.8 | |
| MiDaSPrior=ImageNet - 1M, training samples=2M2025.12 | 23.6 | 63 | |
| MiDaSReal Training Samples=2M2024.12 | 24.2 | 62.2 | |
| VDAAlignment=affine-invariant2026.05 | 25.6 | 58 | |
| FLARE2025.07 | 31.2 | 51.3 | |
| FLARE2026.07 | 31.2 | 51.3 |