Princeton Vision & Learning Lab Logo

Multi-Layer Depth

Public finished submissions for multi-layer depth estimation.

Rank Submission Quads Trips Pairs
all ↑ mixed ↑ layer 1 ↑ layer 3 ↑ layer 5 ↑ layer 7 ↑ all ↑ mixed ↑ layer 1 ↑ layer 3 ↑ layer 5 ↑ layer 7 ↑ all ↑ mixed ↑ layer 1 ↑ layer 3 ↑ layer 5 ↑ layer 7 ↑
1 SeeGroup [1] 70.10 70.85 72.69 63.74 60.17 60.36 74.90 76.95 77.63 66.18 62.31 61.52 82.69 92.56 87.04 69.14 64.33 69.26
2 Multi-head (DAv2) [2] 61.34 60.23 65.62 54.36 51.99 44.73 70.58 71.91 74.41 60.75 56.42 49.41 82.58 90.41 85.81 72.87 63.99 63.21
3 Index Concat (NeWCRFs) [3] 26.90 28.67 24.05 30.01 29.86 70.72 45.25 52.45 41.42 41.61 40.34 71.82 66.39 80.74 66.72 55.66 52.26 71.36
4 Multi-head (NeWCRFs) [4] 25.32 25.44 24.16 27.60 27.11 42.28 41.65 46.22 39.85 38.39 34.28 47.34 63.97 74.91 65.59 54.22 47.33 58.52
5 Recurrent (NeWCRFs) [5] 23.77 22.97 25.91 20.50 20.34 48.31 40.70 42.01 44.64 30.42 27.73 58.90 62.29 69.40 68.13 48.59 42.41 58.15

Submission Footnotes

[1] SeeGroup. Anonymous.

[2] Multi-head (DAv2). Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation. [paper] [code]

[3] Index Concat (NeWCRFs). Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation. [paper] [code]

[4] Multi-head (NeWCRFs). Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation. [paper] [code]

[5] Recurrent (NeWCRFs). Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation.. [paper] [code]