Efficient Combination of Classifiers for 3D Action Recognition

J 2021

Efficient Combination of Classifiers for 3D Action Recognition

SEDMIDUBSKÝ, Jan and Pavel ZEZULA

Basic information

Original name

Efficient Combination of Classifiers for 3D Action Recognition

Authors

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

Edition

Multimedia Systems, Springer, 2021, 0942-4962

Other information

Language

English

Type of outcome

Článek v odborném periodiku

Field of Study

10200 1.2 Computer and information sciences

Country of publisher

United States of America

Confidentiality degree

není předmětem státního či obchodního tajemství

References:

URL

Impact factor

Impact factor: 2.603

RIV identification code

RIV/00216224:14330/21:00118857

Organization unit

Faculty of Informatics

DOI

http://dx.doi.org/10.1007/s00530-021-00767-9

UT WoS

000628724200001

Keywords in English

action recognition;skeleton sequence;fusion;augmentation;normalization

Abstract

V originále

The popular task of 3D human action recognition is almost exclusively solved by training deep-learning classifiers. To achieve high recognition accuracy, input 3D actions are often pre-processed by various normalization or augmentation techniques. However, it is not computationally feasible to train a classifier for each possible variant of training data in order to select the best-performing combination of pre-processing techniques for a given dataset. In this paper, we propose an evaluation procedure that determines the best combination in a very efficient way. In particular, we only train one independent classifier for each available pre-processing technique and estimate the accuracy of a specific combination by efficient fusion of the corresponding classification results based on a strict majority vote rule. In addition, for the best-ranked combination, we can retrospectively apply the normalized/augmented variants of input data to train only a single classifier. This enables to decide whether it is generally better to train a single model, or rather a set of independent classifiers whose results are fused within the classification phase. We evaluate the experiments on single-subject as well as person-interaction datasets of 3D skeleton sequences and all combinations of up to 16 normalization and augmentation techniques, some of them also proposed in this paper.

Links

GA19-02033S, research and development project

Name: Vyhledávání, analytika a anotace datových toků lidských pohybů

Investor: Czech Science Foundation

Citovat

SEDMIDUBSKÝ, Jan and Pavel ZEZULA. Efficient Combination of Classifiers for 3D Action Recognition. Multimedia Systems. Springer, 2021, vol. 27, No 5, p. 941-952. ISSN 0942-4962. Available from: https://dx.doi.org/10.1007/s00530-021-00767-9.

@article{1740896,
   author = {Sedmidubský, Jan and Zezula, Pavel},
   article_number = {5},
   doi = {http://dx.doi.org/10.1007/s00530-021-00767-9},
   keywords = {action recognition;skeleton sequence;fusion;augmentation;normalization},
   language = {eng},
   issn = {0942-4962},
   journal = {Multimedia Systems},
   title = {Efficient Combination of Classifiers for 3D Action Recognition},
   url = {https://link.springer.com/article/10.1007/s00530-021-00767-9},
   volume = {27},
   year = {2021}
}

TY  - JOUR
ID  - 1740896
AU  - Sedmidubský, Jan - Zezula, Pavel
PY  - 2021
TI  - Efficient Combination of Classifiers for 3D Action Recognition
JF  - Multimedia Systems
VL  - 27
IS  - 5
SP  - 941-952
EP  - 941-952
PB  - Springer
SN  - 09424962
KW  - action recognition;skeleton sequence;fusion;augmentation;normalization
UR  - https://link.springer.com/article/10.1007/s00530-021-00767-9
N2  - The popular task of 3D human action recognition is almost exclusively solved by training deep-learning classifiers. To achieve high recognition accuracy, input 3D actions are often pre-processed by various normalization or augmentation techniques. However, it is not computationally feasible to train a classifier for each possible variant of training data in order to select the best-performing combination of pre-processing techniques for a given dataset. In this paper, we propose an evaluation procedure that determines the best combination in a very efficient way. In particular, we only train one independent classifier for each available pre-processing technique and estimate the accuracy of a specific combination by efficient fusion of the corresponding classification results based on a strict majority vote rule. In addition, for the best-ranked combination, we can retrospectively apply the normalized/augmented variants of input data to train only a single classifier. This enables to decide whether it is generally better to train a single model, or rather a set of independent classifiers whose results are fused within the classification phase. We evaluate the experiments on single-subject as well as person-interaction datasets of 3D skeleton sequences and all combinations of up to 16 normalization and augmentation techniques, some of them also proposed in this paper.
ER  -

SEDMIDUBSKÝ, Jan and Pavel ZEZULA. Efficient Combination of Classifiers for 3D Action Recognition. \textit{Multimedia Systems}. Springer, 2021, vol.~27, No~5, p.~941-952. ISSN~0942-4962. Available from: https://dx.doi.org/10.1007/s00530-021-00767-9.

Detailed Information on Publication Record