Method

End-to-end Learning of Cost-Volume Aggregation for Real-time Dense Stereo [DeepCostAggr]
https://github.com/andrey-kuzmin/deep-cost-aggr

Submitted on 19 Nov. 2016 09:59 by
Andrey Kuzmin (Skolkovo Institute of Science and technology)

Running time:0.03 s
Environment:GPU @ 2.5 Ghz (C/C++)

Method Description:
We present a new deep learning-based approach for
dense stereo matching. Compared to previous works, our
approach does not use deep learning of pixel appearance
descriptors, employing very fast classical matching
scores instead. At the same time, our approach uses a
deep convolutional network to predict the local
parameters of cost volume aggregation process, which in
this paper we implement using differentiable domain
transform. By treating such transform as a recurrent
neural network, we are able to train our whole system that
includes cost volume computation, cost-volume
aggregation (smoothing), and winner-takes-all disparity
selection end-to-end. The resulting method is highly
efficient at test time while achieving good matching
accuracy. On the KITTI 2015 benchmark, it achieves a
result of 6.34\% error rate while running at 29 frames per
second rate on a modern GPU.
Parameters:
SAD 1x1, CENSUS 7x7
\alpha = 0.43
\sigma = 4
Latex Bibtex:
@inproceedings{kuzmin2017end,
title={End-to-end Learning of Cost-Volume Aggregation
for
Real-time Dense Stereo},
author={Kuzmin, Andrey and Mikushin, Dmitry and
Lempitsky,
Victor},
booktitle = {2017 IEEE 27th International Workshop on
Machine Learning for Signal Processing (MLSP)},
year={2017}
}

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 5.34 11.35 6.34
All / Est 5.32 11.35 6.32
Noc / All 4.82 10.11 5.69
Noc / Est 4.81 10.11 5.68
This table as LaTeX

Test Image 0

Error D1-bg D1-fg D1-all
All / All 4.97 1.38 4.47
All / Est 4.97 1.38 4.47
Noc / All 4.91 1.38 4.41
Noc / Est 4.91 1.38 4.41
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 1

Error D1-bg D1-fg D1-all
All / All 4.30 6.64 4.56
All / Est 4.14 6.64 4.42
Noc / All 4.01 6.64 4.30
Noc / Est 3.84 6.64 4.16
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 2

Error D1-bg D1-fg D1-all
All / All 5.68 9.09 5.84
All / Est 5.68 9.09 5.84
Noc / All 5.00 9.09 5.20
Noc / Est 5.00 9.09 5.20
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 3

Error D1-bg D1-fg D1-all
All / All 5.92 9.14 6.21
All / Est 5.92 9.14 6.21
Noc / All 5.45 9.14 5.80
Noc / Est 5.45 9.14 5.80
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 4

Error D1-bg D1-fg D1-all
All / All 7.48 4.70 7.02
All / Est 7.48 4.70 7.02
Noc / All 6.48 4.70 6.18
Noc / Est 6.48 4.70 6.18
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 5

Error D1-bg D1-fg D1-all
All / All 11.32 4.20 10.68
All / Est 11.24 4.20 10.61
Noc / All 9.58 4.20 9.08
Noc / Est 9.57 4.20 9.08
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 6

Error D1-bg D1-fg D1-all
All / All 10.17 3.37 9.45
All / Est 10.17 3.37 9.45
Noc / All 9.47 3.37 8.81
Noc / Est 9.47 3.37 8.81
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 7

Error D1-bg D1-fg D1-all
All / All 2.40 8.30 3.55
All / Est 2.39 8.30 3.55
Noc / All 2.41 8.30 3.58
Noc / Est 2.41 8.30 3.58
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 8

Error D1-bg D1-fg D1-all
All / All 2.93 2.31 2.82
All / Est 2.87 2.31 2.77
Noc / All 2.93 2.31 2.82
Noc / Est 2.87 2.31 2.77
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 9

Error D1-bg D1-fg D1-all
All / All 2.34 2.94 2.49
All / Est 2.27 2.94 2.44
Noc / All 2.36 2.82 2.47
Noc / Est 2.29 2.82 2.42
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 10

Error D1-bg D1-fg D1-all
All / All 3.47 7.28 4.34
All / Est 3.47 7.28 4.34
Noc / All 3.44 7.28 4.33
Noc / Est 3.44 7.28 4.33
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 11

Error D1-bg D1-fg D1-all
All / All 2.88 2.69 2.85
All / Est 2.88 2.69 2.84
Noc / All 2.87 2.69 2.84
Noc / Est 2.87 2.69 2.83
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 12

Error D1-bg D1-fg D1-all
All / All 1.51 2.67 1.59
All / Est 1.51 2.67 1.59
Noc / All 1.45 2.67 1.54
Noc / Est 1.45 2.67 1.54
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 13

Error D1-bg D1-fg D1-all
All / All 1.79 2.14 1.84
All / Est 1.79 2.14 1.84
Noc / All 1.69 2.14 1.74
Noc / Est 1.69 2.14 1.74
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 14

Error D1-bg D1-fg D1-all
All / All 2.11 4.55 2.15
All / Est 2.11 4.55 2.15
Noc / All 1.98 4.55 2.02
Noc / Est 1.98 4.55 2.02
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 15

Error D1-bg D1-fg D1-all
All / All 5.16 3.20 4.98
All / Est 5.16 3.20 4.98
Noc / All 5.05 3.20 4.88
Noc / Est 5.05 3.20 4.88
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 16

Error D1-bg D1-fg D1-all
All / All 6.95 3.30 6.41
All / Est 6.95 3.30 6.41
Noc / All 6.37 3.30 5.91
Noc / Est 6.37 3.30 5.91
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 17

Error D1-bg D1-fg D1-all
All / All 2.82 1.73 2.71
All / Est 2.82 1.73 2.71
Noc / All 2.55 1.73 2.47
Noc / Est 2.55 1.73 2.47
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 18

Error D1-bg D1-fg D1-all
All / All 8.43 17.30 12.65
All / Est 8.42 17.30 12.64
Noc / All 7.91 17.30 12.42
Noc / Est 7.91 17.30 12.42
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 19

Error D1-bg D1-fg D1-all
All / All 1.90 4.69 2.22
All / Est 1.90 4.69 2.22
Noc / All 1.79 4.69 2.12
Noc / Est 1.79 4.69 2.12
This table as LaTeX

Input Image

D1 Result

D1 Error




eXTReMe Tracker