A hybrid CNN approach for single image depth estimation: A case study

Károly Harsányi, Attila Kiss, András Majdik, T. Szirányi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Three-dimensional scene understanding is an emerging field in many real-world applications. Autonomous driving, robotics, and continuous real-time tracking are hot topics within the engineering society. One essential component of this is to develop faster and more reliable algorithms being capable of predicting depths from RGB images. Generally, it is easier to install a system with fewer cameras because it requires less calibration. Thus, our aim is to develop a strategy for predicting the depth on a single image as precisely as possible from one point of view. There are existing methods for this problem with promising results. The goal of this paper is to advance the state-of-the-art in the field of single-image depth prediction using convolutional neural networks. In order to do so, we modified an existing deep neural network to get improved results. The proposed architecture contains additional side-to-side connections between the encoding and decoding branches.

LanguageEnglish
Title of host publicationMultimedia and Network Information Systems - Proceedings of the 11th International Conference MISSI 2018
EditorsKazimierz Choros, Marek Kopel, Elzbieta Kukla, Andrzej Sieminski
PublisherSpringer Verlag
Pages372-381
Number of pages10
ISBN (Print)9783319986777
DOIs
Publication statusPublished - Jan 1 2019
Event11th International Conference on Multimedia and Network Information Systems, MISSI 2018 - Wroclaw, Poland
Duration: Sep 12 2018Sep 14 2018

Publication series

NameAdvances in Intelligent Systems and Computing
Volume833
ISSN (Print)2194-5357

Other

Other11th International Conference on Multimedia and Network Information Systems, MISSI 2018
CountryPoland
CityWroclaw
Period9/12/189/14/18

Fingerprint

Decoding
Robotics
Cameras
Calibration
Neural networks
Deep neural networks

Keywords

  • CNN
  • Deep learning
  • Depth estimation

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Computer Science(all)

Cite this

Harsányi, K., Kiss, A., Majdik, A., & Szirányi, T. (2019). A hybrid CNN approach for single image depth estimation: A case study. In K. Choros, M. Kopel, E. Kukla, & A. Sieminski (Eds.), Multimedia and Network Information Systems - Proceedings of the 11th International Conference MISSI 2018 (pp. 372-381). (Advances in Intelligent Systems and Computing; Vol. 833). Springer Verlag. https://doi.org/10.1007/978-3-319-98678-4_38

A hybrid CNN approach for single image depth estimation : A case study. / Harsányi, Károly; Kiss, Attila; Majdik, András; Szirányi, T.

Multimedia and Network Information Systems - Proceedings of the 11th International Conference MISSI 2018. ed. / Kazimierz Choros; Marek Kopel; Elzbieta Kukla; Andrzej Sieminski. Springer Verlag, 2019. p. 372-381 (Advances in Intelligent Systems and Computing; Vol. 833).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Harsányi, K, Kiss, A, Majdik, A & Szirányi, T 2019, A hybrid CNN approach for single image depth estimation: A case study. in K Choros, M Kopel, E Kukla & A Sieminski (eds), Multimedia and Network Information Systems - Proceedings of the 11th International Conference MISSI 2018. Advances in Intelligent Systems and Computing, vol. 833, Springer Verlag, pp. 372-381, 11th International Conference on Multimedia and Network Information Systems, MISSI 2018, Wroclaw, Poland, 9/12/18. https://doi.org/10.1007/978-3-319-98678-4_38
Harsányi K, Kiss A, Majdik A, Szirányi T. A hybrid CNN approach for single image depth estimation: A case study. In Choros K, Kopel M, Kukla E, Sieminski A, editors, Multimedia and Network Information Systems - Proceedings of the 11th International Conference MISSI 2018. Springer Verlag. 2019. p. 372-381. (Advances in Intelligent Systems and Computing). https://doi.org/10.1007/978-3-319-98678-4_38
Harsányi, Károly ; Kiss, Attila ; Majdik, András ; Szirányi, T. / A hybrid CNN approach for single image depth estimation : A case study. Multimedia and Network Information Systems - Proceedings of the 11th International Conference MISSI 2018. editor / Kazimierz Choros ; Marek Kopel ; Elzbieta Kukla ; Andrzej Sieminski. Springer Verlag, 2019. pp. 372-381 (Advances in Intelligent Systems and Computing).
@inproceedings{ef9bcc71644e48b0b6a1a054f41b2d33,
title = "A hybrid CNN approach for single image depth estimation: A case study",
abstract = "Three-dimensional scene understanding is an emerging field in many real-world applications. Autonomous driving, robotics, and continuous real-time tracking are hot topics within the engineering society. One essential component of this is to develop faster and more reliable algorithms being capable of predicting depths from RGB images. Generally, it is easier to install a system with fewer cameras because it requires less calibration. Thus, our aim is to develop a strategy for predicting the depth on a single image as precisely as possible from one point of view. There are existing methods for this problem with promising results. The goal of this paper is to advance the state-of-the-art in the field of single-image depth prediction using convolutional neural networks. In order to do so, we modified an existing deep neural network to get improved results. The proposed architecture contains additional side-to-side connections between the encoding and decoding branches.",
keywords = "CNN, Deep learning, Depth estimation",
author = "K{\'a}roly Hars{\'a}nyi and Attila Kiss and Andr{\'a}s Majdik and T. Szir{\'a}nyi",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-319-98678-4_38",
language = "English",
isbn = "9783319986777",
series = "Advances in Intelligent Systems and Computing",
publisher = "Springer Verlag",
pages = "372--381",
editor = "Kazimierz Choros and Marek Kopel and Elzbieta Kukla and Andrzej Sieminski",
booktitle = "Multimedia and Network Information Systems - Proceedings of the 11th International Conference MISSI 2018",

}

TY - GEN

T1 - A hybrid CNN approach for single image depth estimation

T2 - A case study

AU - Harsányi, Károly

AU - Kiss, Attila

AU - Majdik, András

AU - Szirányi, T.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Three-dimensional scene understanding is an emerging field in many real-world applications. Autonomous driving, robotics, and continuous real-time tracking are hot topics within the engineering society. One essential component of this is to develop faster and more reliable algorithms being capable of predicting depths from RGB images. Generally, it is easier to install a system with fewer cameras because it requires less calibration. Thus, our aim is to develop a strategy for predicting the depth on a single image as precisely as possible from one point of view. There are existing methods for this problem with promising results. The goal of this paper is to advance the state-of-the-art in the field of single-image depth prediction using convolutional neural networks. In order to do so, we modified an existing deep neural network to get improved results. The proposed architecture contains additional side-to-side connections between the encoding and decoding branches.

AB - Three-dimensional scene understanding is an emerging field in many real-world applications. Autonomous driving, robotics, and continuous real-time tracking are hot topics within the engineering society. One essential component of this is to develop faster and more reliable algorithms being capable of predicting depths from RGB images. Generally, it is easier to install a system with fewer cameras because it requires less calibration. Thus, our aim is to develop a strategy for predicting the depth on a single image as precisely as possible from one point of view. There are existing methods for this problem with promising results. The goal of this paper is to advance the state-of-the-art in the field of single-image depth prediction using convolutional neural networks. In order to do so, we modified an existing deep neural network to get improved results. The proposed architecture contains additional side-to-side connections between the encoding and decoding branches.

KW - CNN

KW - Deep learning

KW - Depth estimation

UR - http://www.scopus.com/inward/record.url?scp=85053825222&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053825222&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-98678-4_38

DO - 10.1007/978-3-319-98678-4_38

M3 - Conference contribution

SN - 9783319986777

T3 - Advances in Intelligent Systems and Computing

SP - 372

EP - 381

BT - Multimedia and Network Information Systems - Proceedings of the 11th International Conference MISSI 2018

A2 - Choros, Kazimierz

A2 - Kopel, Marek

A2 - Kukla, Elzbieta

A2 - Sieminski, Andrzej

PB - Springer Verlag

ER -