Princeton Vision & Learning Lab Logo

First Layer Depth

Public finished submissions for first-layer depth estimation.

Rank Submission Pair-wise Accuracy ↑ Triplet-wise Accuracy ↑ Quadruplet-wise Accuracy ↑
1 Metric3D V2 ft. [1] 89.53 81.71 75.20
2 Commercial_NvDepthAnythingV2_ft [2] 88.07 78.42 74.09
3 DepthPro [3] 87.39 76.29 69.46
4 Depth Anything V2 [4] 85.34 74.44 70.43
5 Marigold [5] 82.59 68.35 55.89
6 GeoWizard [6] 81.39 66.29 52.43
7 Metric3D V2 [7] 80.31 65.43 55.14
8 Depth Anything [8] 78.02 62.95 58.88
9 UniDepth V2 [9] 77.03 62.15 56.86
10 MoGe [10] 76.76 63.99 58.92
11 MiDaS v3.1 [11] 76.61 62.05 58.54
12 ZoeDepth [12] 74.25 58.56 52.73

Submission Footnotes

[1] Metric3D V2 ft.. Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation. [paper]

[2] Commercial_NvDepthAnythingV2_ft. Commercial NvDepthAnythingV2 ft. [paper] [code]

[3] DepthPro. Depth Pro: Sharp Monocular Metric Depth in Less Than a Second. [paper] [code]

[4] Depth Anything V2. Depth Anything V2. [paper] [code]

[5] Marigold. Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation. [paper] [code]

[6] GeoWizard. GeoWizard: Unleashing the Diffusion Priors for 3D Geometry Estimation from a Single Image. [paper] [code]

[7] Metric3D V2. Metric3D V2: A Versatile Monocular Geometric Foundation Model for Zero-shot Metric Depth. [paper] [code]

[8] Depth Anything. Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data. [paper] [code]

[9] UniDepth V2. UniDepthV2: Universal Monocular Metric Depth Estimation Made Simpler. [paper] [code]

[10] MoGe. MoGe: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training. [paper] [code]

[11] MiDaS v3.1. MiDaS v3.1 – A Model Zoo for Robust Monocular Relative Depth Estimation. [paper] [code]

[12] ZoeDepth. ZoeDepth: Combining relative and metric depth. [paper] [code]