Depth Map Decomposition for Monocular Depth Estimation

Jinyoung Jun; Jae Han Lee; Chul Lee; Chang Su Kim

doi:10.1007/978-3-031-20086-1_2

Depth Map Decomposition for Monocular Depth Estimation

Jinyoung Jun, Jae Han Lee, Chul Lee, Chang Su Kim

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

8 Citations (Scopus)

Abstract

We propose a novel algorithm for monocular depth estimation that decomposes a metric depth map into a normalized depth map and scale features. The proposed network is composed of a shared encoder and three decoders, called G-Net, N-Net, and M-Net, which estimate gradient maps, a normalized depth map, and a metric depth map, respectively. M-Net learns to estimate metric depths more accurately using relative depth features extracted by G-Net and N-Net. The proposed algorithm has the advantage that it can use datasets without metric depth labels to improve the performance of metric depth estimation. Experimental results on various datasets demonstrate that the proposed algorithm not only provides competitive performance to state-of-the-art algorithms but also yields acceptable results even when only a small amount of metric depth data is available for its training.

Original language	English
Title of host publication	Computer Vision – ECCV 2022 - 17th European Conference, Proceedings
Editors	Shai Avidan, Gabriel Brostow, Moustapha Cissé, Giovanni Maria Farinella, Tal Hassner
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	18-34
Number of pages	17
ISBN (Print)	9783031200854
DOIs	https://doi.org/10.1007/978-3-031-20086-1_2
Publication status	Published - 2022
Event	17th European Conference on Computer Vision, ECCV 2022 - Tel Aviv, Israel Duration: 2022 Oct 23 → 2022 Oct 27

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13662 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	17th European Conference on Computer Vision, ECCV 2022
Country/Territory	Israel
City	Tel Aviv
Period	22/10/23 → 22/10/27

Bibliographical note

Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Keywords

Depth map decomposition
Monocular depth estimation
Relative depth estimation

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-031-20086-1_2

Cite this

Jun, J., Lee, J. H., Lee, C., & Kim, C. S. (2022). Depth Map Decomposition for Monocular Depth Estimation. In S. Avidan, G. Brostow, M. Cissé, G. M. Farinella, & T. Hassner (Eds.), Computer Vision – ECCV 2022 - 17th European Conference, Proceedings (pp. 18-34). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13662 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-20086-1_2

Depth Map Decomposition for Monocular Depth Estimation. / Jun, Jinyoung; Lee, Jae Han; Lee, Chul et al.
Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. ed. / Shai Avidan; Gabriel Brostow; Moustapha Cissé; Giovanni Maria Farinella; Tal Hassner. Springer Science and Business Media Deutschland GmbH, 2022. p. 18-34 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13662 LNCS).

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

Jun, J, Lee, JH, Lee, C & Kim, CS 2022, Depth Map Decomposition for Monocular Depth Estimation. in S Avidan, G Brostow, M Cissé, GM Farinella & T Hassner (eds), Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13662 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 18-34, 17th European Conference on Computer Vision, ECCV 2022, Tel Aviv, Israel, 22/10/23. https://doi.org/10.1007/978-3-031-20086-1_2

Jun J, Lee JH, Lee C, Kim CS. Depth Map Decomposition for Monocular Depth Estimation. In Avidan S, Brostow G, Cissé M, Farinella GM, Hassner T, editors, Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. Springer Science and Business Media Deutschland GmbH. 2022. p. 18-34. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-20086-1_2

Jun, Jinyoung ; Lee, Jae Han ; Lee, Chul et al. / Depth Map Decomposition for Monocular Depth Estimation. Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. editor / Shai Avidan ; Gabriel Brostow ; Moustapha Cissé ; Giovanni Maria Farinella ; Tal Hassner. Springer Science and Business Media Deutschland GmbH, 2022. pp. 18-34 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{ac116b61911844199dc447f474da94de,

title = "Depth Map Decomposition for Monocular Depth Estimation",

abstract = "We propose a novel algorithm for monocular depth estimation that decomposes a metric depth map into a normalized depth map and scale features. The proposed network is composed of a shared encoder and three decoders, called G-Net, N-Net, and M-Net, which estimate gradient maps, a normalized depth map, and a metric depth map, respectively. M-Net learns to estimate metric depths more accurately using relative depth features extracted by G-Net and N-Net. The proposed algorithm has the advantage that it can use datasets without metric depth labels to improve the performance of metric depth estimation. Experimental results on various datasets demonstrate that the proposed algorithm not only provides competitive performance to state-of-the-art algorithms but also yields acceptable results even when only a small amount of metric depth data is available for its training.",

keywords = "Depth map decomposition, Monocular depth estimation, Relative depth estimation",

author = "Jinyoung Jun and Lee, {Jae Han} and Chul Lee and Kim, {Chang Su}",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 17th European Conference on Computer Vision, ECCV 2022 ; Conference date: 23-10-2022 Through 27-10-2022",

year = "2022",

doi = "10.1007/978-3-031-20086-1_2",

language = "English",

isbn = "9783031200854",

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

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "18--34",

editor = "Shai Avidan and Gabriel Brostow and Moustapha Ciss{\'e} and Farinella, {Giovanni Maria} and Tal Hassner",

booktitle = "Computer Vision – ECCV 2022 - 17th European Conference, Proceedings",

address = "Germany",

}

TY - GEN

T1 - Depth Map Decomposition for Monocular Depth Estimation

AU - Jun, Jinyoung

AU - Lee, Jae Han

AU - Lee, Chul

AU - Kim, Chang Su

PY - 2022

Y1 - 2022

N2 - We propose a novel algorithm for monocular depth estimation that decomposes a metric depth map into a normalized depth map and scale features. The proposed network is composed of a shared encoder and three decoders, called G-Net, N-Net, and M-Net, which estimate gradient maps, a normalized depth map, and a metric depth map, respectively. M-Net learns to estimate metric depths more accurately using relative depth features extracted by G-Net and N-Net. The proposed algorithm has the advantage that it can use datasets without metric depth labels to improve the performance of metric depth estimation. Experimental results on various datasets demonstrate that the proposed algorithm not only provides competitive performance to state-of-the-art algorithms but also yields acceptable results even when only a small amount of metric depth data is available for its training.

AB - We propose a novel algorithm for monocular depth estimation that decomposes a metric depth map into a normalized depth map and scale features. The proposed network is composed of a shared encoder and three decoders, called G-Net, N-Net, and M-Net, which estimate gradient maps, a normalized depth map, and a metric depth map, respectively. M-Net learns to estimate metric depths more accurately using relative depth features extracted by G-Net and N-Net. The proposed algorithm has the advantage that it can use datasets without metric depth labels to improve the performance of metric depth estimation. Experimental results on various datasets demonstrate that the proposed algorithm not only provides competitive performance to state-of-the-art algorithms but also yields acceptable results even when only a small amount of metric depth data is available for its training.

KW - Depth map decomposition

KW - Monocular depth estimation

KW - Relative depth estimation

UR - http://www.scopus.com/inward/record.url?scp=85142670452&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-20086-1_2

DO - 10.1007/978-3-031-20086-1_2

M3 - Conference contribution

AN - SCOPUS:85142670452

SN - 9783031200854

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

SP - 18

EP - 34

BT - Computer Vision – ECCV 2022 - 17th European Conference, Proceedings

A2 - Avidan, Shai

A2 - Brostow, Gabriel

A2 - Cissé, Moustapha

A2 - Farinella, Giovanni Maria

A2 - Hassner, Tal

PB - Springer Science and Business Media Deutschland GmbH

T2 - 17th European Conference on Computer Vision, ECCV 2022

Y2 - 23 October 2022 through 27 October 2022

ER -

Depth Map Decomposition for Monocular Depth Estimation

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this