Method

Deep learning based stereo matching method using sing-view videos [SMV]
[Anonymous Submission]

Submitted on 28 Jan. 2019 11:45 by
[Anonymous Submission]

Running time:1.6 min
Environment:8 cores @ 3.5 Ghz (Python)

Method Description:
This paper proposes an unsupervised approach to construct a deep learning based stereo matching method using sing-view videos (SMV). From videos, a set of corresponding points are computed between images, and image patches that center at the computed points are extracted. Negative and positive samples constitute a dataset to train a similarity network that is then used as a matching cost function.

In addition, we propose a local-global matching cost network that exploits the first feature maps (local features) accompanying with last feature features (global features) as output feature of the proposed network. The concatenated features are connected to full-connected layers and the network outputs a similarity measure of an image patch pair as a matching cost. Computed matching costs are aggregated using semi-global matching and cross-based cost aggregation, followed by sub-pixel interpolation, left-right consistency check, median and bilateral filtering.

We evaluate the proposed stereo matching methods using popular stereo matching datasets, including KITTI 2012 and 2015, and Middlebury. We submit the disparity maps to their benchmark servers to evaluate the performance of SMV. We also compared the generalization of SMV and baseline methods using the training sets of the three datasets.
Parameters:
Network Parameters
input_patch_size=11x11
average_pooling_size=7x7
ckernel_size=3
num_clayers=5
num_fc_layers=3
num_fmaps=112
num_fc_units=384
Latex Bibtex:

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 5.03 16.34 6.91
All / Est 5.02 16.34 6.91
Noc / All 4.40 14.89 6.13
Noc / Est 4.39 14.89 6.12
This table as LaTeX

Test Image 0

Error D1-bg D1-fg D1-all
All / All 4.46 3.76 4.37
All / Est 4.46 3.76 4.37
Noc / All 4.30 3.76 4.22
Noc / Est 4.30 3.76 4.22
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 1

Error D1-bg D1-fg D1-all
All / All 3.92 8.30 4.41
All / Est 3.92 8.30 4.41
Noc / All 3.47 8.30 4.02
Noc / Est 3.47 8.30 4.02
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 2

Error D1-bg D1-fg D1-all
All / All 4.93 10.65 5.21
All / Est 4.93 10.65 5.21
Noc / All 3.78 10.65 4.12
Noc / Est 3.78 10.65 4.12
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 3

Error D1-bg D1-fg D1-all
All / All 6.97 33.42 9.41
All / Est 6.96 33.42 9.40
Noc / All 6.12 33.42 8.68
Noc / Est 6.12 33.42 8.68
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 4

Error D1-bg D1-fg D1-all
All / All 6.91 15.52 8.34
All / Est 6.91 15.52 8.34
Noc / All 5.99 15.52 7.60
Noc / Est 5.99 15.52 7.60
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 5

Error D1-bg D1-fg D1-all
All / All 11.40 12.04 11.46
All / Est 11.39 12.04 11.45
Noc / All 9.29 12.04 9.54
Noc / Est 9.28 12.04 9.53
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 6

Error D1-bg D1-fg D1-all
All / All 11.65 4.04 10.85
All / Est 11.65 4.04 10.85
Noc / All 10.71 4.04 10.00
Noc / Est 10.71 4.04 10.00
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 7

Error D1-bg D1-fg D1-all
All / All 2.37 11.99 4.25
All / Est 2.37 11.99 4.25
Noc / All 2.39 11.99 4.29
Noc / Est 2.39 11.99 4.29
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 8

Error D1-bg D1-fg D1-all
All / All 2.42 3.35 2.59
All / Est 2.42 3.35 2.59
Noc / All 2.39 3.35 2.57
Noc / Est 2.39 3.35 2.57
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 9

Error D1-bg D1-fg D1-all
All / All 2.40 7.13 3.61
All / Est 2.40 7.13 3.61
Noc / All 2.42 6.01 3.32
Noc / Est 2.42 6.01 3.32
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 10

Error D1-bg D1-fg D1-all
All / All 3.32 7.49 4.27
All / Est 3.32 7.49 4.27
Noc / All 3.35 7.49 4.31
Noc / Est 3.35 7.49 4.31
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 11

Error D1-bg D1-fg D1-all
All / All 3.26 3.24 3.26
All / Est 3.26 3.24 3.26
Noc / All 3.22 3.24 3.22
Noc / Est 3.22 3.24 3.22
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 12

Error D1-bg D1-fg D1-all
All / All 1.37 5.35 1.64
All / Est 1.37 5.35 1.64
Noc / All 1.32 5.35 1.60
Noc / Est 1.32 5.35 1.60
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 13

Error D1-bg D1-fg D1-all
All / All 1.70 0.56 1.56
All / Est 1.69 0.56 1.55
Noc / All 1.51 0.56 1.39
Noc / Est 1.50 0.56 1.38
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 14

Error D1-bg D1-fg D1-all
All / All 1.54 6.24 1.62
All / Est 1.54 6.24 1.62
Noc / All 1.30 6.24 1.38
Noc / Est 1.30 6.24 1.38
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 15

Error D1-bg D1-fg D1-all
All / All 4.04 7.13 4.32
All / Est 4.04 7.13 4.32
Noc / All 3.71 7.13 4.03
Noc / Est 3.71 7.13 4.03
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 16

Error D1-bg D1-fg D1-all
All / All 6.43 4.56 6.15
All / Est 6.43 4.56 6.15
Noc / All 5.70 4.56 5.53
Noc / Est 5.70 4.56 5.53
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 17

Error D1-bg D1-fg D1-all
All / All 2.33 1.82 2.28
All / Est 2.33 1.82 2.28
Noc / All 1.67 1.82 1.69
Noc / Est 1.67 1.82 1.69
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 18

Error D1-bg D1-fg D1-all
All / All 7.59 19.03 13.03
All / Est 7.59 19.03 13.03
Noc / All 6.88 19.03 12.71
Noc / Est 6.88 19.03 12.71
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 19

Error D1-bg D1-fg D1-all
All / All 1.77 4.02 2.03
All / Est 1.77 4.02 2.03
Noc / All 1.53 4.02 1.82
Noc / Est 1.53 4.02 1.81
This table as LaTeX

Input Image

D1 Result

D1 Error




eXTReMe Tracker