Method

Learning to Aggregate Costs from Multiple Scanline Optimizations in Semi-Global Matching [SGM-Forest]


Submitted on 9 Mar. 2018 15:00 by
Johannes Schönberger (ETH Zürich)

Running time:6 seconds
Environment:1 core @ 3.0 Ghz (Python/C/C++)

Method Description:
Semi-Global Matching (SGM) uses an aggregation scheme to combine costs from multiple 1D scanline optimizations that tends to hurt its accuracy in difficult scenarios. We propose replacing this aggregation scheme with a new learning-based method that fuses disparity proposals estimated using scanline optimization. Our proposed SGM-Forest algorithm solves this problem using per-pixel classification. SGM-Forest currently ranks 1st on the ETH3D stereo benchmark and is ranked competitively on the Middlebury 2014 and KITTI 2015 benchmarks. It consistently outperforms SGM in challenging settings and under difficult training protocols that demonstrate robust generalization, while adding only a small computational overhead to SGM.
Parameters:
Latex Bibtex:
@inproceedings{schoenberger2018sgm,
author={Schönberger, Johannes Lutz and Sinha, Sudipta and Pollefeys, Marc},
title={{Learning to Fuse Proposals from Multiple Scanline Optimizations in Semi-Global Matching}},
booktitle={European Conference on Computer Vision (ECCV)},
year={2018},
}

Detailed Results

This page provides detailed results for the method(s) selected. For the first 20 test images, the percentage of erroneous pixels is depicted in the table. We use the error metric described in Object Scene Flow for Autonomous Vehicles (CVPR 2015), which considers a pixel to be correctly estimated if the disparity or flow end-point error is <3px or <5% (for scene flow this criterion needs to be fulfilled for both disparity maps and the flow map). Underneath, the left input image, the estimated results and the error maps are shown (for disp_0/disp_1/flow/scene_flow, respectively). The error map uses the log-color scale described in Object Scene Flow for Autonomous Vehicles (CVPR 2015), depicting correct estimates (<3px or <5% error) in blue and wrong estimates in red color tones. Dark regions in the error images denote the occluded pixels which fall outside the image boundaries. The false color maps of the results are scaled to the largest ground truth disparity values / flow magnitudes.

Test Set Average

Error D1-bg D1-fg D1-all
All / All 3.11 10.74 4.38
All / Est 3.06 10.58 4.31
Noc / All 2.79 9.70 3.93
Noc / Est 2.78 9.66 3.92
This table as LaTeX

Test Image 0

Error D1-bg D1-fg D1-all
All / All 2.29 2.56 2.33
All / Est 2.26 2.56 2.30
Noc / All 2.26 2.56 2.30
Noc / Est 2.23 2.56 2.28
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 1

Error D1-bg D1-fg D1-all
All / All 2.33 5.74 2.71
All / Est 2.32 5.72 2.70
Noc / All 2.23 5.74 2.63
Noc / Est 2.22 5.72 2.62
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 2

Error D1-bg D1-fg D1-all
All / All 3.94 6.01 4.04
All / Est 3.94 6.01 4.04
Noc / All 3.43 6.01 3.56
Noc / Est 3.43 6.01 3.56
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 3

Error D1-bg D1-fg D1-all
All / All 3.67 7.76 4.05
All / Est 3.67 7.76 4.05
Noc / All 3.25 7.76 3.67
Noc / Est 3.25 7.76 3.67
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 4

Error D1-bg D1-fg D1-all
All / All 5.46 5.74 5.50
All / Est 5.45 5.74 5.50
Noc / All 4.70 5.74 4.88
Noc / Est 4.69 5.74 4.87
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 5

Error D1-bg D1-fg D1-all
All / All 9.41 6.82 9.18
All / Est 9.41 6.82 9.17
Noc / All 7.87 6.82 7.78
Noc / Est 7.87 6.82 7.77
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 6

Error D1-bg D1-fg D1-all
All / All 7.58 3.86 7.19
All / Est 7.47 3.86 7.09
Noc / All 7.47 3.86 7.09
Noc / Est 7.36 3.86 6.99
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 7

Error D1-bg D1-fg D1-all
All / All 1.22 4.11 1.79
All / Est 1.22 4.11 1.79
Noc / All 1.24 4.11 1.81
Noc / Est 1.24 4.11 1.81
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 8

Error D1-bg D1-fg D1-all
All / All 1.32 3.45 1.72
All / Est 1.32 3.45 1.72
Noc / All 1.32 3.45 1.71
Noc / Est 1.32 3.45 1.71
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 9

Error D1-bg D1-fg D1-all
All / All 1.73 2.56 1.94
All / Est 1.73 2.56 1.94
Noc / All 1.75 2.58 1.96
Noc / Est 1.75 2.58 1.96
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 10

Error D1-bg D1-fg D1-all
All / All 1.60 3.63 2.06
All / Est 1.60 3.63 2.06
Noc / All 1.61 3.63 2.08
Noc / Est 1.61 3.63 2.08
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 11

Error D1-bg D1-fg D1-all
All / All 1.34 0.84 1.25
All / Est 1.34 0.84 1.25
Noc / All 1.34 0.84 1.25
Noc / Est 1.34 0.84 1.25
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 12

Error D1-bg D1-fg D1-all
All / All 0.72 2.15 0.82
All / Est 0.72 2.15 0.82
Noc / All 0.72 2.15 0.82
Noc / Est 0.72 2.15 0.82
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 13

Error D1-bg D1-fg D1-all
All / All 0.94 0.96 0.94
All / Est 0.93 0.96 0.94
Noc / All 0.78 0.96 0.80
Noc / Est 0.77 0.96 0.79
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 14

Error D1-bg D1-fg D1-all
All / All 1.44 2.19 1.45
All / Est 1.44 2.19 1.45
Noc / All 1.34 2.19 1.35
Noc / Est 1.34 2.19 1.35
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 15

Error D1-bg D1-fg D1-all
All / All 2.99 1.99 2.90
All / Est 2.99 1.99 2.90
Noc / All 3.02 1.99 2.93
Noc / Est 3.02 1.99 2.93
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 16

Error D1-bg D1-fg D1-all
All / All 5.43 1.25 4.81
All / Est 5.42 1.25 4.80
Noc / All 5.02 1.25 4.46
Noc / Est 5.01 1.25 4.45
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 17

Error D1-bg D1-fg D1-all
All / All 1.61 0.67 1.51
All / Est 1.61 0.67 1.51
Noc / All 1.40 0.67 1.32
Noc / Est 1.40 0.67 1.32
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 18

Error D1-bg D1-fg D1-all
All / All 5.97 11.00 8.36
All / Est 5.97 11.00 8.36
Noc / All 5.79 11.00 8.29
Noc / Est 5.79 11.00 8.29
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 19

Error D1-bg D1-fg D1-all
All / All 1.11 3.09 1.33
All / Est 1.10 3.09 1.32
Noc / All 1.09 3.09 1.32
Noc / Est 1.08 3.09 1.31
This table as LaTeX

Input Image

D1 Result

D1 Error




eXTReMe Tracker