EdgeConv with Attention Module for Monocular Depth Estimation

Minhyeok Lee; Sangwon Hwang; Chaewon Park; Sangyoun Lee

doi:10.1109/WACV51458.2022.00242

EdgeConv with Attention Module for Monocular Depth Estimation

Minhyeok Lee, Sangwon Hwang, Chaewon Park, Sangyoun Lee

Department of Electrical and Electronic Engineering

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

Abstract

Monocular depth estimation is an especially important task in robotics and autonomous driving, where 3D structural information is essential. However, extreme lighting conditions and complex surface objects make it difficult to predict depth in a single image. Therefore, to generate accurate depth maps, it is important for the model to learn structural information about the scene. We propose a novel Patch-Wise EdgeConv Module (PEM) and EdgeConv Attention Module (EAM) to solve the difficulty of monocular depth estimation. The proposed modules extract structural information by learning the relationship between image patches close to each other in space using edge convolution. Our method is evaluated on two popular datasets, the NYU Depth V2 and the KITTI Eigen split, achieving state-of-the-art performance. We prove that the proposed model predicts depth robustly in challenging scenes through various comparative experiments.

Original language	English
Title of host publication	Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	2364-2373
Number of pages	10
ISBN (Electronic)	9781665409155
DOIs	https://doi.org/10.1109/WACV51458.2022.00242
Publication status	Published - 2022
Event	22nd IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022 - Waikoloa, United States Duration: 2022 Jan 4 → 2022 Jan 8

Publication series

Name	Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022

Conference

Conference	22nd IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022
Country/Territory	United States
City	Waikoloa
Period	22/1/4 → 22/1/8

Bibliographical note

Funding Information:
Acknowledgement. This work was supported by the Institute of Information & communications Technology Planning & Evaluation(IITP) grant funded by the Korea government(MSIT) (No. 2021-0-00172, The development of human Re-identification and masked face recognition based on CCTV camera)

Publisher Copyright:
© 2022 IEEE.

All Science Journal Classification (ASJC) codes

Computer Vision and Pattern Recognition
Computer Science Applications

Access to Document

10.1109/WACV51458.2022.00242

Cite this

Lee, M., Hwang, S., Park, C., & Lee, S. (2022). EdgeConv with Attention Module for Monocular Depth Estimation. In Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022 (pp. 2364-2373). (Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/WACV51458.2022.00242

Lee, Minhyeok ; Hwang, Sangwon ; Park, Chaewon ; Lee, Sangyoun. / EdgeConv with Attention Module for Monocular Depth Estimation. Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022. Institute of Electrical and Electronics Engineers Inc., 2022. pp. 2364-2373 (Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022).

@inproceedings{de371f74cc5c48eb81ad8564b7637628,

title = "EdgeConv with Attention Module for Monocular Depth Estimation",

abstract = "Monocular depth estimation is an especially important task in robotics and autonomous driving, where 3D structural information is essential. However, extreme lighting conditions and complex surface objects make it difficult to predict depth in a single image. Therefore, to generate accurate depth maps, it is important for the model to learn structural information about the scene. We propose a novel Patch-Wise EdgeConv Module (PEM) and EdgeConv Attention Module (EAM) to solve the difficulty of monocular depth estimation. The proposed modules extract structural information by learning the relationship between image patches close to each other in space using edge convolution. Our method is evaluated on two popular datasets, the NYU Depth V2 and the KITTI Eigen split, achieving state-of-the-art performance. We prove that the proposed model predicts depth robustly in challenging scenes through various comparative experiments. ",

author = "Minhyeok Lee and Sangwon Hwang and Chaewon Park and Sangyoun Lee",

note = "Funding Information: Acknowledgement. This work was supported by the Institute of Information & communications Technology Planning & Evaluation(IITP) grant funded by the Korea government(MSIT) (No. 2021-0-00172, The development of human Re-identification and masked face recognition based on CCTV camera) Publisher Copyright: {\textcopyright} 2022 IEEE.; 22nd IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022 ; Conference date: 04-01-2022 Through 08-01-2022",

year = "2022",

doi = "10.1109/WACV51458.2022.00242",

language = "English",

series = "Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "2364--2373",

booktitle = "Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022",

address = "United States",

}

Lee, M, Hwang, S, Park, C & Lee, S 2022, EdgeConv with Attention Module for Monocular Depth Estimation. in Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022. Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022, Institute of Electrical and Electronics Engineers Inc., pp. 2364-2373, 22nd IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022, Waikoloa, United States, 22/1/4. https://doi.org/10.1109/WACV51458.2022.00242

EdgeConv with Attention Module for Monocular Depth Estimation. / Lee, Minhyeok; Hwang, Sangwon; Park, Chaewon; Lee, Sangyoun.

Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022. Institute of Electrical and Electronics Engineers Inc., 2022. p. 2364-2373 (Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022).

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

TY - GEN

T1 - EdgeConv with Attention Module for Monocular Depth Estimation

AU - Lee, Minhyeok

AU - Hwang, Sangwon

AU - Park, Chaewon

AU - Lee, Sangyoun

N1 - Funding Information: Acknowledgement. This work was supported by the Institute of Information & communications Technology Planning & Evaluation(IITP) grant funded by the Korea government(MSIT) (No. 2021-0-00172, The development of human Re-identification and masked face recognition based on CCTV camera) Publisher Copyright: © 2022 IEEE.

PY - 2022

Y1 - 2022

N2 - Monocular depth estimation is an especially important task in robotics and autonomous driving, where 3D structural information is essential. However, extreme lighting conditions and complex surface objects make it difficult to predict depth in a single image. Therefore, to generate accurate depth maps, it is important for the model to learn structural information about the scene. We propose a novel Patch-Wise EdgeConv Module (PEM) and EdgeConv Attention Module (EAM) to solve the difficulty of monocular depth estimation. The proposed modules extract structural information by learning the relationship between image patches close to each other in space using edge convolution. Our method is evaluated on two popular datasets, the NYU Depth V2 and the KITTI Eigen split, achieving state-of-the-art performance. We prove that the proposed model predicts depth robustly in challenging scenes through various comparative experiments.

AB - Monocular depth estimation is an especially important task in robotics and autonomous driving, where 3D structural information is essential. However, extreme lighting conditions and complex surface objects make it difficult to predict depth in a single image. Therefore, to generate accurate depth maps, it is important for the model to learn structural information about the scene. We propose a novel Patch-Wise EdgeConv Module (PEM) and EdgeConv Attention Module (EAM) to solve the difficulty of monocular depth estimation. The proposed modules extract structural information by learning the relationship between image patches close to each other in space using edge convolution. Our method is evaluated on two popular datasets, the NYU Depth V2 and the KITTI Eigen split, achieving state-of-the-art performance. We prove that the proposed model predicts depth robustly in challenging scenes through various comparative experiments.

UR - http://www.scopus.com/inward/record.url?scp=85126143596&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85126143596&partnerID=8YFLogxK

U2 - 10.1109/WACV51458.2022.00242

DO - 10.1109/WACV51458.2022.00242

M3 - Conference contribution

AN - SCOPUS:85126143596

T3 - Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022

SP - 2364

EP - 2373

BT - Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 22nd IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022

Y2 - 4 January 2022 through 8 January 2022

ER -

Lee M, Hwang S, Park C, Lee S. EdgeConv with Attention Module for Monocular Depth Estimation. In Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022. Institute of Electrical and Electronics Engineers Inc. 2022. p. 2364-2373. (Proceedings - 2022 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2022). https://doi.org/10.1109/WACV51458.2022.00242

EdgeConv with Attention Module for Monocular Depth Estimation

Abstract

Publication series

Conference

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this