Princeton Vision & Learning Lab Logo

ZEDD Leaderboard

All public finished submissions are listed below.

Rank Submission d1.05 ↑ d1.15 ↑ d1.25 & d1 ↑ AbsRel ↓
1 FOSSA-vitb [1] 0.504756 0.823506 0.917954 0.089240
2 FOSSA-vits [2] 0.450551 0.803226 0.884313 0.097731
3 DepthPro [3] 0.182310 0.488849 0.665051 0.200530
4 DFF-FV [4] 0.152989 0.410359 0.575589 0.368678
5 DFF-DFV [5] 0.152032 0.391168 0.546482 0.573026
6 MoGe-2 [6] 0.148703 0.376841 0.579867 0.277563
7 DEReD [7] 0.128098 0.356129 0.504990 0.347281
8 UniDepthv2 [8] 0.105566 0.399752 0.604524 0.252958
9 HybridDepth [9] 0.075192 0.230750 0.346785 0.754496
10 DepthAnythingV2-ViTS-Indoors [10] 0.036770 0.162068 0.261067 0.336901

Submission Footnotes

[1] FOSSA-vitb. Zero-Shot Depth from Defocus. [website link] [paper link]

[2] FOSSA-vits. Zero-Shot Depth from Defocus. [website link] [paper link]

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

[4] DFF-FV. Deep Depth from Focus with Differential Focus Volume. [website link] [paper link]

[5] DFF-DFV. Deep Depth from Focus with Differential Focus Volume. [website link] [paper link]

[6] MoGe-2. MoGe-2: Accurate Monocular Geometry with Metric Scale and Sharp Details. [website link] [paper link]

[7] DEReD. Fully Self-Supervised Depth Estimation from Defocus Clue. [website link] [paper link]

[8] UniDepthv2. UniDepthV2: Universal Monocular Metric Depth Estimation Made Simpler. [website link] [paper link]

[9] HybridDepth. Hybrid Depth: Robust Depth Fusion By Leveraging Depth from Focus and Single-Image Priors. [website link] [paper link]

[10] DepthAnythingV2-ViTS-Indoors. Depth Anything V2. [website link] [paper link]