Strengthening Skeletal Action Recognizers via Leveraging Temporal Patterns

Zhenyue Qin; Pan Ji; Dongwoo Kim; Yang Liu; Saeed Anwar; Tom Gedeon

doi:10.1007/978-3-031-25072-9_39

Strengthening Skeletal Action Recognizers via Leveraging Temporal Patterns

Zhenyue Qin^*
, Pan Ji
, Dongwoo Kim
, Yang Liu
, Saeed Anwar
, Tom Gedeon

^*Corresponding author for this work

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

2 Scopus citations

Abstract

Skeleton sequences are compact and lightweight. Numerous skeleton-based action recognizers have been proposed to classify human behaviors. In this work, we aim to incorporate components that are compatible with existing models and further improve their accuracy. To this end, we design two temporal accessories: discrete cosine encoding (DCE) and chronological loss (CRL). DCE facilitates models to analyze motion patterns from the frequency domain and meanwhile alleviates the influence of signal noise. CRL guides networks to explicitly capture the sequence’s chronological order. These two components consistently endow many recently-proposed action recognizers with accuracy boosts, achieving new state-of-the-art (SOTA) accuracy on two large datasets.

Original language	English
Title of host publication	Computer Vision – ECCV 2022 Workshops, Proceedings
Editors	Leonid Karlinsky, Tomer Michaeli, Ko Nishino
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	577-593
Number of pages	17
ISBN (Print)	9783031250712
DOIs	https://doi.org/10.1007/978-3-031-25072-9_39
State	Published - 2023
Externally published	Yes
Event	Workshops held at the 17th European Conference on Computer Vision, ECCV 2022 - Tel Aviv, Israel Duration: 23 Oct 2022 → 27 Oct 2022

Publication series

Name	Lecture Notes in Computer Science
Volume	13805 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	Workshops held at the 17th European Conference on Computer Vision, ECCV 2022
Country/Territory	Israel
City	Tel Aviv
Period	23/10/22 → 27/10/22

Bibliographical note

Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-031-25072-9_39

Cite this

Qin, Z., Ji, P., Kim, D., Liu, Y., Anwar, S., & Gedeon, T. (2023). Strengthening Skeletal Action Recognizers via Leveraging Temporal Patterns. In L. Karlinsky, T. Michaeli, & K. Nishino (Eds.), Computer Vision – ECCV 2022 Workshops, Proceedings (pp. 577-593). (Lecture Notes in Computer Science; Vol. 13805 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-25072-9_39

@inproceedings{53678179c93e44ef88aabe8f4883a66d,

title = "Strengthening Skeletal Action Recognizers via Leveraging Temporal Patterns",

abstract = "Skeleton sequences are compact and lightweight. Numerous skeleton-based action recognizers have been proposed to classify human behaviors. In this work, we aim to incorporate components that are compatible with existing models and further improve their accuracy. To this end, we design two temporal accessories: discrete cosine encoding (DCE) and chronological loss (CRL). DCE facilitates models to analyze motion patterns from the frequency domain and meanwhile alleviates the influence of signal noise. CRL guides networks to explicitly capture the sequence{\textquoteright}s chronological order. These two components consistently endow many recently-proposed action recognizers with accuracy boosts, achieving new state-of-the-art (SOTA) accuracy on two large datasets.",

author = "Zhenyue Qin and Pan Ji and Dongwoo Kim and Yang Liu and Saeed Anwar and Tom Gedeon",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; Workshops held at the 17th European Conference on Computer Vision, ECCV 2022 ; Conference date: 23-10-2022 Through 27-10-2022",

year = "2023",

doi = "10.1007/978-3-031-25072-9\_39",

language = "English",

isbn = "9783031250712",

series = "Lecture Notes in Computer Science",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "577--593",

editor = "Leonid Karlinsky and Tomer Michaeli and Ko Nishino",

booktitle = "Computer Vision – ECCV 2022 Workshops, Proceedings",

address = "Germany",

}

Qin, Z, Ji, P, Kim, D, Liu, Y, Anwar, S & Gedeon, T 2023, Strengthening Skeletal Action Recognizers via Leveraging Temporal Patterns. in L Karlinsky, T Michaeli & K Nishino (eds), Computer Vision – ECCV 2022 Workshops, Proceedings. Lecture Notes in Computer Science, vol. 13805 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 577-593, Workshops held at the 17th European Conference on Computer Vision, ECCV 2022, Tel Aviv, Israel, 23/10/22. https://doi.org/10.1007/978-3-031-25072-9_39

Strengthening Skeletal Action Recognizers via Leveraging Temporal Patterns. / Qin, Zhenyue; Ji, Pan; Kim, Dongwoo et al.
Computer Vision – ECCV 2022 Workshops, Proceedings. ed. / Leonid Karlinsky; Tomer Michaeli; Ko Nishino. Springer Science and Business Media Deutschland GmbH, 2023. p. 577-593 (Lecture Notes in Computer Science; Vol. 13805 LNCS).

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

TY - GEN

T1 - Strengthening Skeletal Action Recognizers via Leveraging Temporal Patterns

AU - Qin, Zhenyue

AU - Ji, Pan

AU - Kim, Dongwoo

AU - Liu, Yang

AU - Anwar, Saeed

AU - Gedeon, Tom

PY - 2023

Y1 - 2023

N2 - Skeleton sequences are compact and lightweight. Numerous skeleton-based action recognizers have been proposed to classify human behaviors. In this work, we aim to incorporate components that are compatible with existing models and further improve their accuracy. To this end, we design two temporal accessories: discrete cosine encoding (DCE) and chronological loss (CRL). DCE facilitates models to analyze motion patterns from the frequency domain and meanwhile alleviates the influence of signal noise. CRL guides networks to explicitly capture the sequence’s chronological order. These two components consistently endow many recently-proposed action recognizers with accuracy boosts, achieving new state-of-the-art (SOTA) accuracy on two large datasets.

AB - Skeleton sequences are compact and lightweight. Numerous skeleton-based action recognizers have been proposed to classify human behaviors. In this work, we aim to incorporate components that are compatible with existing models and further improve their accuracy. To this end, we design two temporal accessories: discrete cosine encoding (DCE) and chronological loss (CRL). DCE facilitates models to analyze motion patterns from the frequency domain and meanwhile alleviates the influence of signal noise. CRL guides networks to explicitly capture the sequence’s chronological order. These two components consistently endow many recently-proposed action recognizers with accuracy boosts, achieving new state-of-the-art (SOTA) accuracy on two large datasets.

UR - https://www.scopus.com/pages/publications/85150963439

U2 - 10.1007/978-3-031-25072-9_39

DO - 10.1007/978-3-031-25072-9_39

M3 - Conference contribution

AN - SCOPUS:85150963439

SN - 9783031250712

T3 - Lecture Notes in Computer Science

SP - 577

EP - 593

BT - Computer Vision – ECCV 2022 Workshops, Proceedings

A2 - Karlinsky, Leonid

A2 - Michaeli, Tomer

A2 - Nishino, Ko

PB - Springer Science and Business Media Deutschland GmbH

T2 - Workshops held at the 17th European Conference on Computer Vision, ECCV 2022

Y2 - 23 October 2022 through 27 October 2022

ER -