Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement

Minhyeok Heo; Jaehan Lee; Kyung Rae Kim; Han Ul Kim; Chang-Su Kim

doi:10.1007/978-3-030-01225-0_3

Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement

Minhyeok Heo, Jaehan Lee, Kyung Rae Kim, Han Ul Kim, Chang-Su Kim

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Citation (Scopus)

Abstract

We propose a monocular depth estimation algorithm based on whole strip masking (WSM) and reliability-based refinement. First, we develop a convolutional neural network (CNN) tailored for the depth estimation. Specifically, we design a novel filter, called WSM, to exploit the tendency that a scene has similar depths in horizonal or vertical directions. The proposed CNN combines WSM upsampling blocks with a ResNet encoder. Second, we measure the reliability of an estimated depth, by appending additional layers to the main CNN. Using the reliability information, we perform conditional random field (CRF) optimization to refine the estimated depth map. Experimental results demonstrate that the proposed algorithm provides the state-of-the-art depth estimation performance.

Original language	English
Title of host publication	Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings
Editors	Vittorio Ferrari, Cristian Sminchisescu, Yair Weiss, Martial Hebert
Publisher	Springer Verlag
Pages	39-55
Number of pages	17
ISBN (Print)	9783030012243
DOIs	https://doi.org/10.1007/978-3-030-01225-0_3
Publication status	Published - 2018 Jan 1
Event	15th European Conference on Computer Vision, ECCV 2018 - Munich, Germany Duration: 2018 Sep 8 → 2018 Sep 14

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11208 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	15th European Conference on Computer Vision, ECCV 2018
Country	Germany
City	Munich
Period	18/9/8 → 18/9/14

Keywords

Depth map refinement
Monocular depth estimation
Reliability
Whole strip masking

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-030-01225-0_3

Fingerprint Dive into the research topics of 'Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement'. Together they form a unique fingerprint.

Cite this

Heo, M., Lee, J., Kim, K. R., Kim, H. U., & Kim, C-S. (2018). Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement. In V. Ferrari, C. Sminchisescu, Y. Weiss, & M. Hebert (Eds.), Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings (pp. 39-55). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11208 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-01225-0_3

Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement. / Heo, Minhyeok; Lee, Jaehan; Kim, Kyung Rae; Kim, Han Ul; Kim, Chang-Su.

Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. ed. / Vittorio Ferrari; Cristian Sminchisescu; Yair Weiss; Martial Hebert. Springer Verlag, 2018. p. 39-55 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11208 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Heo, M, Lee, J, Kim, KR, Kim, HU & Kim, C-S 2018, Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement. in V Ferrari, C Sminchisescu, Y Weiss & M Hebert (eds), Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11208 LNCS, Springer Verlag, pp. 39-55, 15th European Conference on Computer Vision, ECCV 2018, Munich, Germany, 18/9/8. https://doi.org/10.1007/978-3-030-01225-0_3

Heo M, Lee J, Kim KR, Kim HU, Kim C-S. Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement. In Ferrari V, Sminchisescu C, Weiss Y, Hebert M, editors, Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. Springer Verlag. 2018. p. 39-55. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-01225-0_3

Heo, Minhyeok ; Lee, Jaehan ; Kim, Kyung Rae ; Kim, Han Ul ; Kim, Chang-Su. / Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement. Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. editor / Vittorio Ferrari ; Cristian Sminchisescu ; Yair Weiss ; Martial Hebert. Springer Verlag, 2018. pp. 39-55 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{5a6261f26e7b4addadb26f98cf4e314d,

title = "Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement",

abstract = "We propose a monocular depth estimation algorithm based on whole strip masking (WSM) and reliability-based refinement. First, we develop a convolutional neural network (CNN) tailored for the depth estimation. Specifically, we design a novel filter, called WSM, to exploit the tendency that a scene has similar depths in horizonal or vertical directions. The proposed CNN combines WSM upsampling blocks with a ResNet encoder. Second, we measure the reliability of an estimated depth, by appending additional layers to the main CNN. Using the reliability information, we perform conditional random field (CRF) optimization to refine the estimated depth map. Experimental results demonstrate that the proposed algorithm provides the state-of-the-art depth estimation performance.",

keywords = "Depth map refinement, Monocular depth estimation, Reliability, Whole strip masking",

author = "Minhyeok Heo and Jaehan Lee and Kim, {Kyung Rae} and Kim, {Han Ul} and Chang-Su Kim",

year = "2018",

month = jan,

day = "1",

doi = "10.1007/978-3-030-01225-0_3",

language = "English",

isbn = "9783030012243",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "39--55",

editor = "Vittorio Ferrari and Cristian Sminchisescu and Yair Weiss and Martial Hebert",

booktitle = "Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings",

note = "15th European Conference on Computer Vision, ECCV 2018 ; Conference date: 08-09-2018 Through 14-09-2018",

}

TY - GEN

T1 - Monocular Depth Estimation Using Whole Strip Masking and Reliability-Based Refinement

AU - Heo, Minhyeok

AU - Lee, Jaehan

AU - Kim, Kyung Rae

AU - Kim, Han Ul

AU - Kim, Chang-Su

PY - 2018/1/1

Y1 - 2018/1/1

N2 - We propose a monocular depth estimation algorithm based on whole strip masking (WSM) and reliability-based refinement. First, we develop a convolutional neural network (CNN) tailored for the depth estimation. Specifically, we design a novel filter, called WSM, to exploit the tendency that a scene has similar depths in horizonal or vertical directions. The proposed CNN combines WSM upsampling blocks with a ResNet encoder. Second, we measure the reliability of an estimated depth, by appending additional layers to the main CNN. Using the reliability information, we perform conditional random field (CRF) optimization to refine the estimated depth map. Experimental results demonstrate that the proposed algorithm provides the state-of-the-art depth estimation performance.

AB - We propose a monocular depth estimation algorithm based on whole strip masking (WSM) and reliability-based refinement. First, we develop a convolutional neural network (CNN) tailored for the depth estimation. Specifically, we design a novel filter, called WSM, to exploit the tendency that a scene has similar depths in horizonal or vertical directions. The proposed CNN combines WSM upsampling blocks with a ResNet encoder. Second, we measure the reliability of an estimated depth, by appending additional layers to the main CNN. Using the reliability information, we perform conditional random field (CRF) optimization to refine the estimated depth map. Experimental results demonstrate that the proposed algorithm provides the state-of-the-art depth estimation performance.

KW - Depth map refinement

KW - Monocular depth estimation

KW - Reliability

KW - Whole strip masking

UR - http://www.scopus.com/inward/record.url?scp=85055438092&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85055438092&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-01225-0_3

DO - 10.1007/978-3-030-01225-0_3

M3 - Conference contribution

AN - SCOPUS:85055438092

SN - 9783030012243

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 39

EP - 55

BT - Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings

A2 - Ferrari, Vittorio

A2 - Sminchisescu, Cristian

A2 - Weiss, Yair

A2 - Hebert, Martial

PB - Springer Verlag

T2 - 15th European Conference on Computer Vision, ECCV 2018

Y2 - 8 September 2018 through 14 September 2018

ER -