Method

Bilateral Grid Learning for Stereo Matching Network [BGNet]


Submitted on 16 Nov. 2020 12:56 by
Yuhua Xu (Hefei University of Technology)

Running time:0.02 s
Environment:GPU @ >3.5 Ghz (Python)

Method Description:
We present a novel edge-preserving cost volume
upsampling module based on the slicing operation
in the learned bilateral grid. The slicing layer
is parameter-free, which allows us to obtain a
high quality cost volume of high resolution from a
low resolution cost volume under the guide of the
learned guidance map efficiently. The proposed
cost volume upsampling module can be seamlessly
embedded into many existing stereo matching
networks, such as GCNet, PSMNet, and GANet. The
resulting networks are accelerated several times
while maintaining comparable accuracy.
Furthermore, based on this module we design a
real-time network (named BGNet).
Parameters:
Adam, bata1 = 0.9, beta2 = 0.999, learning rate =
0.001
Latex Bibtex:
@article{bgnet2021,
title={Bilateral Grid Learning for Stereo
Matching Network},
author={Xu, Bin and Xu, Yuhua and Yang, Xiaoli
and Jia, Wei and Guo, Yulan},
journal={Proceedings of the IEEE/CVF Conference
on Computer Vision and Pattern Recognition (CVPR)},
year={2021}
}

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 2.07 4.74 2.51
All / Est 2.07 4.74 2.51
Noc / All 1.91 4.34 2.31
Noc / Est 1.91 4.34 2.31
This table as LaTeX

Test Image 0

Error D1-bg D1-fg D1-all
All / All 2.48 1.23 2.31
All / Est 2.48 1.23 2.31
Noc / All 2.43 1.23 2.27
Noc / Est 2.43 1.23 2.27
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 1

Error D1-bg D1-fg D1-all
All / All 2.25 1.56 2.18
All / Est 2.25 1.56 2.18
Noc / All 2.19 1.56 2.12
Noc / Est 2.19 1.56 2.12
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 2

Error D1-bg D1-fg D1-all
All / All 2.84 4.86 2.94
All / Est 2.84 4.86 2.94
Noc / All 2.58 4.86 2.69
Noc / Est 2.58 4.86 2.69
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 3

Error D1-bg D1-fg D1-all
All / All 3.28 0.57 3.03
All / Est 3.28 0.57 3.03
Noc / All 3.26 0.57 3.01
Noc / Est 3.26 0.57 3.01
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 4

Error D1-bg D1-fg D1-all
All / All 2.01 1.64 1.95
All / Est 2.01 1.64 1.95
Noc / All 1.81 1.64 1.78
Noc / Est 1.81 1.64 1.78
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 5

Error D1-bg D1-fg D1-all
All / All 3.45 2.86 3.40
All / Est 3.45 2.86 3.40
Noc / All 3.26 2.86 3.23
Noc / Est 3.26 2.86 3.23
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 6

Error D1-bg D1-fg D1-all
All / All 4.76 1.82 4.45
All / Est 4.76 1.82 4.45
Noc / All 4.86 1.82 4.54
Noc / Est 4.86 1.82 4.54
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 7

Error D1-bg D1-fg D1-all
All / All 0.80 3.15 1.26
All / Est 0.80 3.15 1.26
Noc / All 0.81 3.15 1.27
Noc / Est 0.81 3.15 1.27
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 8

Error D1-bg D1-fg D1-all
All / All 0.71 1.84 0.92
All / Est 0.71 1.84 0.92
Noc / All 0.70 1.84 0.91
Noc / Est 0.70 1.84 0.91
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 9

Error D1-bg D1-fg D1-all
All / All 0.91 2.07 1.21
All / Est 0.91 2.07 1.21
Noc / All 0.92 2.17 1.23
Noc / Est 0.92 2.17 1.23
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 10

Error D1-bg D1-fg D1-all
All / All 1.22 3.97 1.85
All / Est 1.22 3.97 1.85
Noc / All 1.24 3.97 1.87
Noc / Est 1.24 3.97 1.87
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 11

Error D1-bg D1-fg D1-all
All / All 1.03 0.94 1.01
All / Est 1.03 0.94 1.01
Noc / All 1.03 0.94 1.02
Noc / Est 1.03 0.94 1.02
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 12

Error D1-bg D1-fg D1-all
All / All 0.84 0.45 0.81
All / Est 0.84 0.45 0.81
Noc / All 0.70 0.45 0.68
Noc / Est 0.70 0.45 0.68
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 13

Error D1-bg D1-fg D1-all
All / All 0.85 0.25 0.77
All / Est 0.85 0.25 0.77
Noc / All 0.69 0.25 0.63
Noc / Est 0.69 0.25 0.63
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 14

Error D1-bg D1-fg D1-all
All / All 1.41 0.96 1.40
All / Est 1.41 0.96 1.40
Noc / All 1.30 0.96 1.29
Noc / Est 1.30 0.96 1.29
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 15

Error D1-bg D1-fg D1-all
All / All 3.36 0.75 3.12
All / Est 3.36 0.75 3.12
Noc / All 3.43 0.75 3.18
Noc / Est 3.43 0.75 3.18
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 16

Error D1-bg D1-fg D1-all
All / All 4.38 0.47 3.80
All / Est 4.38 0.47 3.80
Noc / All 4.18 0.47 3.63
Noc / Est 4.18 0.47 3.63
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 17

Error D1-bg D1-fg D1-all
All / All 0.78 0.41 0.75
All / Est 0.78 0.41 0.75
Noc / All 0.78 0.41 0.74
Noc / Est 0.78 0.41 0.74
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 18

Error D1-bg D1-fg D1-all
All / All 5.35 3.48 4.46
All / Est 5.35 3.48 4.46
Noc / All 5.24 3.48 4.39
Noc / Est 5.24 3.48 4.39
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 19

Error D1-bg D1-fg D1-all
All / All 0.92 1.16 0.95
All / Est 0.92 1.16 0.95
Noc / All 0.91 1.16 0.94
Noc / Est 0.91 1.16 0.94
This table as LaTeX

Input Image

D1 Result

D1 Error




eXTReMe Tracker