Princeton Vision & Learning Lab Logo

Multi-Layer Depth Estimation

Task

Under this setting, we evaluate methods' multi-layer depth estimation on full LayeredDepth benchmark.

Metrics

Pair-wise, triplet-wise, and quad-wise accuracy on all, mixed, Layer1, Layer3, Layer5, and Layer7 subset of LayeredDepth benchmark. Refer to the paper for detailed definitions.

Leaderboard


All Mixed Layer1 Layer3 Layer5 Layer7
Method P↑ T↑ Q↑ P↑ T↑ Q↑ P↑ T↑ Q↑ P↑ T↑ Q↑ P↑ T↑ Q↑ P↑ T↑ Q↑
MH (DA v2) [1] 80.72 70.67 61.68 89.37 72.98 62.23 85.80 74.34 65.62 73.75 62.68 59.19 66.19 59.29 54.75 57.39 50.78 48.09
MH (NeWCRFs) [2] 63.42 42.55 25.97 74.72 46.13 26.38 65.66 39.93 24.21 56.58 43.36 30.77 52.17 38.66 29.94 38.40 35.65 33.10
IC (NeWCRFs) [3] 64.46 44.00 26.00 76.70 48.37 26.46 66.95 41.84 24.45 55.85 41.53 30.06 55.36 42.69 31.57 39.66 45.18 45.08
R (NeWCRFs) [4] 62.36 41.88 24.64 73.51 45.27 25.35 68.08 44.46 26.12 49.47 31.53 21.25 45.51 30.06 21.29 30.09 34.10 27.59

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

[2] Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation.

[3] Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation.

[4] Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation.

-->