Knowledge graphs meet geometry for semi-supervised monocular depth estimation

Yu Zhao; Fusheng Jin; Mengyuan Wang; Shuliang Wang

doi:10.1007/978-3-030-55130-8_4

Knowledge graphs meet geometry for semi-supervised monocular depth estimation

Yu Zhao, Fusheng Jin^*, Mengyuan Wang, Shuliang Wang

^*Corresponding author for this work

School of Computer Science and Technology

Beijing Institute of Technology

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

2 Citations (Scopus)

Abstract

Depth estimation from a single image plays an important role in computer vision. Using semantic information for depth estimation becomes a research hotspot. The traditional neural network-based semantic method only divides the image according to the features, and cannot understand the deep background knowledge about the real world. In recent years, the knowledge graph is proposed and used for model semantic knowledge. In this paper, we enhance the traditional depth prediction method by analyzing the semantic information of the image through the knowledge graph. Background knowledge from the knowledge graph is used to enhance the results of semantic segmentation, and further improve the depth estimation results. We conducted experiments on the KITTI driving dataset, and the results showed that our method outperformed the previous unsupervised learning methods and supervised learning methods. The result of the Apollo dataset demonstrates that our method can perform in the common case.

Original language	English
Title of host publication	Knowledge Science, Engineering and Management - 13th International Conference, KSEM 2020, Proceedings, Part 1
Editors	Gang Li, Heng Tao Shen, Ye Yuan, Xiaoyang Wang, Huawen Liu, Xiang Zhao
Publisher	Springer
Pages	40-52
Number of pages	13
ISBN (Print)	9783030551292
DOIs	https://doi.org/10.1007/978-3-030-55130-8_4
Publication status	Published - 2020
Event	13th International Conference on Knowledge Science, Engineering and Management, KSEM 2020 - Hangzhou, China Duration: 28 Aug 2020 → 30 Aug 2020

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12274 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	13th International Conference on Knowledge Science, Engineering and Management, KSEM 2020
Country/Territory	China
City	Hangzhou
Period	28/08/20 → 30/08/20

Keywords

Auto driving
Depth prediction
Knowledge graph
Object detection

Access to Document

10.1007/978-3-030-55130-8_4

Cite this

Zhao, Y., Jin, F., Wang, M., & Wang, S. (2020). Knowledge graphs meet geometry for semi-supervised monocular depth estimation. In G. Li, H. T. Shen, Y. Yuan, X. Wang, H. Liu, & X. Zhao (Eds.), Knowledge Science, Engineering and Management - 13th International Conference, KSEM 2020, Proceedings, Part 1 (pp. 40-52). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12274 LNAI). Springer. https://doi.org/10.1007/978-3-030-55130-8_4

Zhao, Yu ; Jin, Fusheng ; Wang, Mengyuan et al. / Knowledge graphs meet geometry for semi-supervised monocular depth estimation. Knowledge Science, Engineering and Management - 13th International Conference, KSEM 2020, Proceedings, Part 1. editor / Gang Li ; Heng Tao Shen ; Ye Yuan ; Xiaoyang Wang ; Huawen Liu ; Xiang Zhao. Springer, 2020. pp. 40-52 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{573692c291a7467ba624209d087b067a,

title = "Knowledge graphs meet geometry for semi-supervised monocular depth estimation",

abstract = "Depth estimation from a single image plays an important role in computer vision. Using semantic information for depth estimation becomes a research hotspot. The traditional neural network-based semantic method only divides the image according to the features, and cannot understand the deep background knowledge about the real world. In recent years, the knowledge graph is proposed and used for model semantic knowledge. In this paper, we enhance the traditional depth prediction method by analyzing the semantic information of the image through the knowledge graph. Background knowledge from the knowledge graph is used to enhance the results of semantic segmentation, and further improve the depth estimation results. We conducted experiments on the KITTI driving dataset, and the results showed that our method outperformed the previous unsupervised learning methods and supervised learning methods. The result of the Apollo dataset demonstrates that our method can perform in the common case.",

keywords = "Auto driving, Depth prediction, Knowledge graph, Object detection",

author = "Yu Zhao and Fusheng Jin and Mengyuan Wang and Shuliang Wang",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2020.; 13th International Conference on Knowledge Science, Engineering and Management, KSEM 2020 ; Conference date: 28-08-2020 Through 30-08-2020",

year = "2020",

doi = "10.1007/978-3-030-55130-8_4",

language = "English",

isbn = "9783030551292",

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

publisher = "Springer",

pages = "40--52",

editor = "Gang Li and Shen, {Heng Tao} and Ye Yuan and Xiaoyang Wang and Huawen Liu and Xiang Zhao",

booktitle = "Knowledge Science, Engineering and Management - 13th International Conference, KSEM 2020, Proceedings, Part 1",

address = "Germany",

}

Zhao, Y, Jin, F, Wang, M & Wang, S 2020, Knowledge graphs meet geometry for semi-supervised monocular depth estimation. in G Li, HT Shen, Y Yuan, X Wang, H Liu & X Zhao (eds), Knowledge Science, Engineering and Management - 13th International Conference, KSEM 2020, Proceedings, Part 1. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12274 LNAI, Springer, pp. 40-52, 13th International Conference on Knowledge Science, Engineering and Management, KSEM 2020, Hangzhou, China, 28/08/20. https://doi.org/10.1007/978-3-030-55130-8_4

Knowledge graphs meet geometry for semi-supervised monocular depth estimation. / Zhao, Yu; Jin, Fusheng; Wang, Mengyuan et al.
Knowledge Science, Engineering and Management - 13th International Conference, KSEM 2020, Proceedings, Part 1. ed. / Gang Li; Heng Tao Shen; Ye Yuan; Xiaoyang Wang; Huawen Liu; Xiang Zhao. Springer, 2020. p. 40-52 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12274 LNAI).

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

TY - GEN

T1 - Knowledge graphs meet geometry for semi-supervised monocular depth estimation

AU - Zhao, Yu

AU - Jin, Fusheng

AU - Wang, Mengyuan

AU - Wang, Shuliang

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2020.

PY - 2020

Y1 - 2020

N2 - Depth estimation from a single image plays an important role in computer vision. Using semantic information for depth estimation becomes a research hotspot. The traditional neural network-based semantic method only divides the image according to the features, and cannot understand the deep background knowledge about the real world. In recent years, the knowledge graph is proposed and used for model semantic knowledge. In this paper, we enhance the traditional depth prediction method by analyzing the semantic information of the image through the knowledge graph. Background knowledge from the knowledge graph is used to enhance the results of semantic segmentation, and further improve the depth estimation results. We conducted experiments on the KITTI driving dataset, and the results showed that our method outperformed the previous unsupervised learning methods and supervised learning methods. The result of the Apollo dataset demonstrates that our method can perform in the common case.

AB - Depth estimation from a single image plays an important role in computer vision. Using semantic information for depth estimation becomes a research hotspot. The traditional neural network-based semantic method only divides the image according to the features, and cannot understand the deep background knowledge about the real world. In recent years, the knowledge graph is proposed and used for model semantic knowledge. In this paper, we enhance the traditional depth prediction method by analyzing the semantic information of the image through the knowledge graph. Background knowledge from the knowledge graph is used to enhance the results of semantic segmentation, and further improve the depth estimation results. We conducted experiments on the KITTI driving dataset, and the results showed that our method outperformed the previous unsupervised learning methods and supervised learning methods. The result of the Apollo dataset demonstrates that our method can perform in the common case.

KW - Auto driving

KW - Depth prediction

KW - Knowledge graph

KW - Object detection

UR - http://www.scopus.com/inward/record.url?scp=85090097056&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-55130-8_4

DO - 10.1007/978-3-030-55130-8_4

M3 - Conference contribution

AN - SCOPUS:85090097056

SN - 9783030551292

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

SP - 40

EP - 52

BT - Knowledge Science, Engineering and Management - 13th International Conference, KSEM 2020, Proceedings, Part 1

A2 - Li, Gang

A2 - Shen, Heng Tao

A2 - Yuan, Ye

A2 - Wang, Xiaoyang

A2 - Liu, Huawen

A2 - Zhao, Xiang

PB - Springer

T2 - 13th International Conference on Knowledge Science, Engineering and Management, KSEM 2020

Y2 - 28 August 2020 through 30 August 2020

ER -

Zhao Y, Jin F, Wang M, Wang S. Knowledge graphs meet geometry for semi-supervised monocular depth estimation. In Li G, Shen HT, Yuan Y, Wang X, Liu H, Zhao X, editors, Knowledge Science, Engineering and Management - 13th International Conference, KSEM 2020, Proceedings, Part 1. Springer. 2020. p. 40-52. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-55130-8_4

Knowledge graphs meet geometry for semi-supervised monocular depth estimation

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this