Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar

Kubra Tas; Deniz Kumlu; Isin Erer

doi:10.1109/TSP52935.2021.9522613

Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar

Kubra Tas, Deniz Kumlu, Isin Erer

Department of Electronics and Communication Engineering

Istanbul Technical University

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

2 Citations (Scopus)

Abstract

A deep learning-based missing data recovery approach is presented for subsurface images with missing samples. The proposed method is based on Pyramid-context Encoder Network (PEN-Net). With this network, region affinity is captured by creating a high-level semantic feature map, and missing data is recovered in a pyramid fashion, for both visual and semantic consistency. Considering missing data cases during subsurface image acquisition, this study aims to obtain plausible recovered images for possible post-processing operations that can be implemented later. Missing data scenarios are constructed in two ways; column-wise and pixel-wise missing data. Each case is tested under 10%, 30% and 50% of missing data scenarios. Based on the experiments that we conducted, it can be observed that better results are obtained with PEN-Net architecture, compared with low rank missing data recovery methods such as Go Decomposition (GoDec) or Low-rank matrix fitting (LmaFit).

Original language	English
Title of host publication	2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021
Editors	Norbert Herencsar
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	263-266
Number of pages	4
ISBN (Electronic)	9781665429337
DOIs	https://doi.org/10.1109/TSP52935.2021.9522613
Publication status	Published - 26 Jul 2021
Event	44th International Conference on Telecommunications and Signal Processing, TSP 2021 - Virtual, Brno, Czech Republic Duration: 26 Jul 2021 → 28 Jul 2021

Publication series

Name	2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021

Conference

Conference	44th International Conference on Telecommunications and Signal Processing, TSP 2021
Country/Territory	Czech Republic
City	Virtual, Brno
Period	26/07/21 → 28/07/21

Bibliographical note

Publisher Copyright:
© 2021 IEEE.

Funding

VI. ACKNOWLEDGEMENT This work is supported by the Scientific and Technological Research Council of Turkey (TUBITAK) under Project No.120E234.

Funders	Funder number
TUBITAK	120E234
Türkiye Bilimsel ve Teknolojik Araştirma Kurumu

Keywords

deep learning
matrix completion
missing data recovery
subsurface imaging

Access to Document

10.1109/TSP52935.2021.9522613

Cite this

Tas, K., Kumlu, D., & Erer, I. (2021). Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar. In N. Herencsar (Ed.), 2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021 (pp. 263-266). (2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/TSP52935.2021.9522613

Tas, Kubra ; Kumlu, Deniz ; Erer, Isin. / Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar. 2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021. editor / Norbert Herencsar. Institute of Electrical and Electronics Engineers Inc., 2021. pp. 263-266 (2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021).

@inproceedings{30f0995f02984f358e03280e63436fb5,

title = "Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar",

abstract = "A deep learning-based missing data recovery approach is presented for subsurface images with missing samples. The proposed method is based on Pyramid-context Encoder Network (PEN-Net). With this network, region affinity is captured by creating a high-level semantic feature map, and missing data is recovered in a pyramid fashion, for both visual and semantic consistency. Considering missing data cases during subsurface image acquisition, this study aims to obtain plausible recovered images for possible post-processing operations that can be implemented later. Missing data scenarios are constructed in two ways; column-wise and pixel-wise missing data. Each case is tested under 10%, 30% and 50% of missing data scenarios. Based on the experiments that we conducted, it can be observed that better results are obtained with PEN-Net architecture, compared with low rank missing data recovery methods such as Go Decomposition (GoDec) or Low-rank matrix fitting (LmaFit).",

keywords = "deep learning, matrix completion, missing data recovery, subsurface imaging",

author = "Kubra Tas and Deniz Kumlu and Isin Erer",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.; 44th International Conference on Telecommunications and Signal Processing, TSP 2021 ; Conference date: 26-07-2021 Through 28-07-2021",

year = "2021",

month = jul,

day = "26",

doi = "10.1109/TSP52935.2021.9522613",

language = "English",

series = "2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021",

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

pages = "263--266",

editor = "Norbert Herencsar",

booktitle = "2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021",

address = "United States",

}

Tas, K, Kumlu, D & Erer, I 2021, Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar. in N Herencsar (ed.), 2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021. 2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021, Institute of Electrical and Electronics Engineers Inc., pp. 263-266, 44th International Conference on Telecommunications and Signal Processing, TSP 2021, Virtual, Brno, Czech Republic, 26/07/21. https://doi.org/10.1109/TSP52935.2021.9522613

Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar. / Tas, Kubra; Kumlu, Deniz; Erer, Isin.
2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021. ed. / Norbert Herencsar. Institute of Electrical and Electronics Engineers Inc., 2021. p. 263-266 (2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021).

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

TY - GEN

T1 - Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar

AU - Tas, Kubra

AU - Kumlu, Deniz

AU - Erer, Isin

PY - 2021/7/26

Y1 - 2021/7/26

N2 - A deep learning-based missing data recovery approach is presented for subsurface images with missing samples. The proposed method is based on Pyramid-context Encoder Network (PEN-Net). With this network, region affinity is captured by creating a high-level semantic feature map, and missing data is recovered in a pyramid fashion, for both visual and semantic consistency. Considering missing data cases during subsurface image acquisition, this study aims to obtain plausible recovered images for possible post-processing operations that can be implemented later. Missing data scenarios are constructed in two ways; column-wise and pixel-wise missing data. Each case is tested under 10%, 30% and 50% of missing data scenarios. Based on the experiments that we conducted, it can be observed that better results are obtained with PEN-Net architecture, compared with low rank missing data recovery methods such as Go Decomposition (GoDec) or Low-rank matrix fitting (LmaFit).

AB - A deep learning-based missing data recovery approach is presented for subsurface images with missing samples. The proposed method is based on Pyramid-context Encoder Network (PEN-Net). With this network, region affinity is captured by creating a high-level semantic feature map, and missing data is recovered in a pyramid fashion, for both visual and semantic consistency. Considering missing data cases during subsurface image acquisition, this study aims to obtain plausible recovered images for possible post-processing operations that can be implemented later. Missing data scenarios are constructed in two ways; column-wise and pixel-wise missing data. Each case is tested under 10%, 30% and 50% of missing data scenarios. Based on the experiments that we conducted, it can be observed that better results are obtained with PEN-Net architecture, compared with low rank missing data recovery methods such as Go Decomposition (GoDec) or Low-rank matrix fitting (LmaFit).

KW - deep learning

KW - matrix completion

KW - missing data recovery

KW - subsurface imaging

UR - http://www.scopus.com/inward/record.url?scp=85115441756&partnerID=8YFLogxK

U2 - 10.1109/TSP52935.2021.9522613

DO - 10.1109/TSP52935.2021.9522613

M3 - Conference contribution

AN - SCOPUS:85115441756

T3 - 2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021

SP - 263

EP - 266

BT - 2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021

A2 - Herencsar, Norbert

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 44th International Conference on Telecommunications and Signal Processing, TSP 2021

Y2 - 26 July 2021 through 28 July 2021

ER -

Tas K, Kumlu D, Erer I. Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar. In Herencsar N, editor, 2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021. Institute of Electrical and Electronics Engineers Inc. 2021. p. 263-266. (2021 44th International Conference on Telecommunications and Signal Processing, TSP 2021). doi: 10.1109/TSP52935.2021.9522613

Pyramid-Context Encoder Network (PEN-Net) for Missing Data Recovery in Ground Penetrating Radar

Abstract

Publication series

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this