Augmenting Spatio-Temporal Human Motion Data for Effective 3D
Action Recognition

D 2019

Augmenting Spatio-Temporal Human Motion Data for Effective 3D Action Recognition

SEDMIDUBSKÝ, Jan a Pavel ZEZULA

Základní údaje

Originální název

Augmenting Spatio-Temporal Human Motion Data for Effective 3D Action Recognition

Autoři

SEDMIDUBSKÝ, Jan (203 Česká republika, garant, domácí) a Pavel ZEZULA (203 Česká republika, domácí)

Vydání

Neuveden, 21st IEEE International Symposium on Multimedia (ISM), od s. 204-207, 4 s. 2019

Nakladatel

IEEE Computer Society

Další údaje

Jazyk

angličtina

Typ výsledku

Stať ve sborníku

Obor

10200 1.2 Computer and information sciences

Utajení

není předmětem státního či obchodního tajemství

Forma vydání

elektronická verze "online"

Kód RIV

RIV/00216224:14330/19:00107708

Organizační jednotka

Fakulta informatiky

ISBN

978-1-72815-606-4

DOI

http://dx.doi.org/10.1109/ISM46123.2019.00044

UT WoS

000528909200033

Klíčová slova anglicky

3D skeleton sequence;multimedia data;data augmentation;action recognition;bidirectional LSTM

Štítky

DISA, firank_B

Příznaky

Mezinárodní význam, Recenzováno

Změněno: 12. 5. 2020 23:41, RNDr. Pavel Šmerk, Ph.D.

Anotace

V originále

Action recognition is a fundamental operation in 3D human motion analysis. Existing deep learning classifiers achieve a high recognition accuracy if large amounts of training data are provided. However, such data are difficult to obtain in a variety of application scenarios, mainly due to the high costs of motion capture technologies and an absence of suitable actors. In this paper, we propose augmentation techniques to artificially enlarge existing collections of 3D human skeleton sequences. The proposed techniques are especially useful for datasets distinguishing in a high number of classes, each of them characterized by only a limited number of action samples. We experimentally demonstrate that the augmented data help to significantly increase the recognition accuracy even using a standard deep learning architecture.

Návaznosti

GA19-02033S, projekt VaV

Název: Vyhledávání, analytika a anotace datových toků lidských pohybů

Investor: Grantová agentura ČR, Searching, Mining, and Annotating Human Motion Streams

Citovat

SEDMIDUBSKÝ, Jan a Pavel ZEZULA. Augmenting Spatio-Temporal Human Motion Data for Effective 3D Action Recognition. Online. In 21st IEEE International Symposium on Multimedia (ISM). Neuveden: IEEE Computer Society, 2019, s. 204-207. ISBN 978-1-72815-606-4. Dostupné z: https://dx.doi.org/10.1109/ISM46123.2019.00044.

@inproceedings{1566957,
   author = {Sedmidubský, Jan and Zezula, Pavel},
   address = {Neuveden},
   booktitle = {21st IEEE International Symposium on Multimedia (ISM)},
   doi = {http://dx.doi.org/10.1109/ISM46123.2019.00044},
   keywords = {3D skeleton sequence;multimedia data;data augmentation;action recognition;bidirectional LSTM},
   howpublished = {elektronická verze "online"},
   language = {eng},
   location = {Neuveden},
   isbn = {978-1-72815-606-4},
   pages = {204-207},
   publisher = {IEEE Computer Society},
   title = {Augmenting Spatio-Temporal Human Motion Data for Effective 3D Action Recognition},
   year = {2019}
}

TY  - JOUR
ID  - 1566957
AU  - Sedmidubský, Jan - Zezula, Pavel
PY  - 2019
TI  - Augmenting Spatio-Temporal Human Motion Data for Effective 3D Action Recognition
PB  - IEEE Computer Society
CY  - Neuveden
SN  - 9781728156064
KW  - 3D skeleton sequence;multimedia data;data augmentation;action recognition;bidirectional LSTM
N2  - Action recognition is a fundamental operation in 3D human motion analysis. Existing deep learning classifiers achieve a high recognition accuracy if large amounts of training data are provided. However, such data are difficult to obtain in a variety of application scenarios, mainly due to the high costs of motion capture technologies and an absence of suitable actors. In this paper, we propose augmentation techniques to artificially enlarge existing collections of 3D human skeleton sequences. The proposed techniques are especially useful for datasets distinguishing in a high number of classes, each of them characterized by only a limited number of action samples. We experimentally demonstrate that the augmented data help to significantly increase the recognition accuracy even using a standard deep learning architecture.
ER  -

SEDMIDUBSKÝ, Jan a Pavel ZEZULA. Augmenting Spatio-Temporal Human Motion Data for Effective 3D Action Recognition. Online. In \textit{21st IEEE International Symposium on Multimedia (ISM)}. Neuveden: IEEE Computer Society, 2019, s.~204-207. ISBN~978-1-72815-606-4. Dostupné z: https://dx.doi.org/10.1109/ISM46123.2019.00044.

Podrobný výpis o publikaci