Understanding the Gap between 2D and 3D Skeleton-Based Action
Recognition

D 2019

Understanding the Gap between 2D and 3D Skeleton-Based Action Recognition

ELIÁŠ, Petr, Jan SEDMIDUBSKÝ and Pavel ZEZULA

Basic information

Original name

Understanding the Gap between 2D and 3D Skeleton-Based Action Recognition

Authors

ELIÁŠ, Petr (203 Czech Republic, belonging to the institution), Jan SEDMIDUBSKÝ (203 Czech Republic, belonging to the institution) and Pavel ZEZULA (203 Czech Republic, belonging to the institution)

Edition

Neuveden, 21st IEEE International Symposium on Multimedia (ISM), p. 192-195, 4 pp. 2019

Publisher

IEEE Computer Society

Other information

Language

English

Type of outcome

Stať ve sborníku

Field of Study

10200 1.2 Computer and information sciences

Confidentiality degree

není předmětem státního či obchodního tajemství

Publication form

electronic version available online

RIV identification code

RIV/00216224:14330/19:00107709

Organization unit

Faculty of Informatics

ISBN

978-1-72815-606-4

DOI

http://dx.doi.org/10.1109/ISM46123.2019.00041

UT WoS

000528909200030

Keywords in English

2D skeleton data;3D skeleton data;action recognition;LSTM;motion data understanding

Abstract

V originále

Large volumes of RGB video data are recorded and processed every day. One of the embedded data modality within these videos is the information about human motions. Up to now, this information has been almost unfeasible to extract, and thus human-motion understanding research has been mainly limited to 3D skeleton data captured by dedicated hardware only. However, with recent advances in computer vision, it is possible to estimate 2D skeleton sequences from ordinary videos quite accurately. Such 2D skeleton data possess an excellent potential for future motion understanding applications. In this paper, we adopt a state-of-the-art bidirectional LSTM network to analyze the accuracy gap in the expressive power of 2D and 3D skeleton data recorded simultaneously on a high number of 20k human actions. We further examine how the missing depth information and fluctuations in 2D skeleton sizes influence the recognition rate. We also demonstrate the suitability of 2D skeleton data for general daily activity recognition by reporting baselines on the PKU-MMD dataset.

Links

GA19-02033S, research and development project

Name: Vyhledávání, analytika a anotace datových toků lidských pohybů

Investor: Czech Science Foundation

Citovat

ELIÁŠ, Petr, Jan SEDMIDUBSKÝ and Pavel ZEZULA. Understanding the Gap between 2D and 3D Skeleton-Based Action Recognition. Online. In 21st IEEE International Symposium on Multimedia (ISM). Neuveden: IEEE Computer Society, 2019, p. 192-195. ISBN 978-1-72815-606-4. Available from: https://dx.doi.org/10.1109/ISM46123.2019.00041.

@inproceedings{1566959,
   author = {Eliáš, Petr and Sedmidubský, Jan and Zezula, Pavel},
   address = {Neuveden},
   booktitle = {21st IEEE International Symposium on Multimedia (ISM)},
   doi = {http://dx.doi.org/10.1109/ISM46123.2019.00041},
   keywords = {2D skeleton data;3D skeleton data;action recognition;LSTM;motion data understanding},
   howpublished = {elektronická verze "online"},
   language = {eng},
   location = {Neuveden},
   isbn = {978-1-72815-606-4},
   pages = {192-195},
   publisher = {IEEE Computer Society},
   title = {Understanding the Gap between 2D and 3D Skeleton-Based Action Recognition},
   year = {2019}
}

TY  - JOUR
ID  - 1566959
AU  - Eliáš, Petr - Sedmidubský, Jan - Zezula, Pavel
PY  - 2019
TI  - Understanding the Gap between 2D and 3D Skeleton-Based Action Recognition
PB  - IEEE Computer Society
CY  - Neuveden
SN  - 9781728156064
KW  - 2D skeleton data;3D skeleton data;action recognition;LSTM;motion data understanding
N2  - Large volumes of RGB video data are recorded and processed every day. One of the embedded data modality within these videos is the information about human motions. Up to now, this information has been almost unfeasible to extract, and thus human-motion understanding research has been mainly limited to 3D skeleton data captured by dedicated hardware only. However, with recent advances in computer vision, it is possible to estimate 2D skeleton sequences from ordinary videos quite accurately. Such 2D skeleton data possess an excellent potential for future motion understanding applications. In this paper, we adopt a state-of-the-art bidirectional LSTM network to analyze the accuracy gap in the expressive power of 2D and 3D skeleton data recorded simultaneously on a high number of 20k human actions. We further examine how the missing depth information and fluctuations in 2D skeleton sizes influence the recognition rate. We also demonstrate the suitability of 2D skeleton data for general daily activity recognition by reporting baselines on the PKU-MMD dataset.
ER  -

ELIÁŠ, Petr, Jan SEDMIDUBSKÝ and Pavel ZEZULA. Understanding the Gap between 2D and 3D Skeleton-Based Action Recognition. Online. In \textit{21st IEEE International Symposium on Multimedia (ISM)}. Neuveden: IEEE Computer Society, 2019, p.~192-195. ISBN~978-1-72815-606-4. Available from: https://dx.doi.org/10.1109/ISM46123.2019.00041.

Detailed Information on Publication Record