Method

Hierarchical Binary Classification for Monocular Depth Estimation [HBC]


Submitted on 3 Jan. 2020 01:40 by
Hualie Jiang (The Chinese University of Hong Kong, Shenzhen)

Running time:0.05 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
Extracting dense depth from a single image is an
important and challenging task in computer vision.
Recently, significant progress has been made in
monocular depth estimation (MDE) by formulating
this problem as a multi-label classification
instead of traditional regression. However, to
discretize depth into hundreds of labels results
in complicate network output and a large number of
parameters. In this paper, we propose to encode
the discretized depth into a binary code and
formulate MDE as a hierarchical binary
classification (HBC) problem to reduce the
complexity. Besides, by studying the distribution
of individual bit of the encoded depth codes, we
find that our encoding scheme can also solve the
problem of depth data imbalance. We conduct
experiments on KITTI and NYU Depth V2 datasets,
which show that our simplified approach still
achieves a comparable performance with state-of-
the-art methods.
Parameters:
learning rate = 10e-4
Latex Bibtex:
@inproceedings{jiang2019hbc,
title={Hierarchical Binary Classification
for Monocular Depth Estimation},
author={Hualie Jiang and Rui Huang},
booktitle={IEEE International Conference on
Robotics and Biomimetics},
year={2019}
}

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 Sparsity Invariant CNNs (THREEDV 2017), 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 Sparsity Invariant CNNs (THREEDV 2017), 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

SILog sqErrorRel absErrorRel iRMSE
Error 15.18 3.79 12.33 17.86
This table as LaTeX

Test Image 0

SILog sqErrorRel absErrorRel iRMSE
Error 12.30 2.02 9.24 10.83
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 1

SILog sqErrorRel absErrorRel iRMSE
Error 15.88 4.62 12.57 23.06
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 2

SILog sqErrorRel absErrorRel iRMSE
Error 28.02 6.42 20.48 40.62
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 3

SILog sqErrorRel absErrorRel iRMSE
Error 12.94 1.74 9.86 13.96
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 4

SILog sqErrorRel absErrorRel iRMSE
Error 16.09 4.04 10.79 20.04
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 5

SILog sqErrorRel absErrorRel iRMSE
Error 14.10 4.41 18.06 28.60
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 6

SILog sqErrorRel absErrorRel iRMSE
Error 12.67 1.72 8.54 10.67
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 7

SILog sqErrorRel absErrorRel iRMSE
Error 9.65 1.50 8.53 11.95
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 8

SILog sqErrorRel absErrorRel iRMSE
Error 19.39 3.54 13.52 22.12
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 9

SILog sqErrorRel absErrorRel iRMSE
Error 19.07 4.48 13.13 21.22
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 10

SILog sqErrorRel absErrorRel iRMSE
Error 7.95 0.69 5.45 7.43
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 11

SILog sqErrorRel absErrorRel iRMSE
Error 32.70 13.40 25.62 33.37
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 12

SILog sqErrorRel absErrorRel iRMSE
Error 6.74 0.51 5.44 5.16
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 13

SILog sqErrorRel absErrorRel iRMSE
Error 12.29 1.56 7.73 10.64
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 14

SILog sqErrorRel absErrorRel iRMSE
Error 3.39 0.13 1.91 4.10
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 15

SILog sqErrorRel absErrorRel iRMSE
Error 11.04 1.47 7.75 19.85
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 16

SILog sqErrorRel absErrorRel iRMSE
Error 12.17 2.48 8.15 9.01
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 17

SILog sqErrorRel absErrorRel iRMSE
Error 20.50 7.95 18.38 30.40
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 18

SILog sqErrorRel absErrorRel iRMSE
Error 31.61 10.49 23.93 34.94
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 19

SILog sqErrorRel absErrorRel iRMSE
Error 23.17 4.93 16.99 39.24
This table as LaTeX

Input Image

D1 Result

D1 Error




eXTReMe Tracker