MultiAct: Long-Term 3D Human Motion Generation from Multiple Action Labels

Taeryung Lee; Gyeongsik Moon; Kyoung Mu Lee

doi:10.1609/aaai.v37i1.25206

MultiAct: Long-Term 3D Human Motion Generation from Multiple Action Labels

Taeryung Lee^*
, Gyeongsik Moon^*
, Kyoung Mu Lee

^*Corresponding author for this work

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

33 Citations (Scopus)

Abstract

We tackle the problem of generating long-term 3D human motion from multiple action labels. Two main previous approaches, such as action- and motion-conditioned methods, have limitations to solve this problem. The action-conditioned methods generate a sequence of motion from a single action. Hence, it cannot generate long-term motions composed of multiple actions and transitions between actions. Meanwhile, the motion-conditioned methods generate future motions from initial motion. The generated future motions only depend on the past, so they are not controllable by the user’s desired actions. We present MultiAct, the first framework to generate long-term 3D human motion from multiple action labels. MultiAct takes account of both action and motion conditions with a unified recurrent generation system. It repetitively takes the previous motion and action label; then, it generates a smooth transition and the motion of the given action. As a result, MultiAct produces realistic long-term motion controlled by the given sequence of multiple action labels. Code is publicly available in https://github.com/TaeryungLee/MultiAct RELEASE.

Original language	English
Title of host publication	AAAI-23 Technical Tracks 1
Editors	Brian Williams, Yiling Chen, Jennifer Neville
Publisher	AAAI press
Pages	1231-1239
Number of pages	9
ISBN (Electronic)	9781577358800
DOIs	https://doi.org/10.1609/aaai.v37i1.25206
Publication status	Published - 2023 Jun 27
Externally published	Yes
Event	37th AAAI Conference on Artificial Intelligence, AAAI 2023 - Washington, United States Duration: 2023 Feb 7 → 2023 Feb 14

Publication series

Name	Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023
Volume	37

Conference

Conference	37th AAAI Conference on Artificial Intelligence, AAAI 2023
Country/Territory	United States
City	Washington
Period	23/2/7 → 23/2/14

Bibliographical note

Publisher Copyright:
Copyright © 2023, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.

ASJC Scopus subject areas

Artificial Intelligence

Access to Document

10.1609/aaai.v37i1.25206

Cite this

MultiAct: Long-Term 3D Human Motion Generation from Multiple Action Labels. / Lee, Taeryung; Moon, Gyeongsik; Lee, Kyoung Mu.
AAAI-23 Technical Tracks 1. ed. / Brian Williams; Yiling Chen; Jennifer Neville. AAAI press, 2023. p. 1231-1239 (Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023; Vol. 37).

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

Lee, T, Moon, G & Lee, KM 2023, MultiAct: Long-Term 3D Human Motion Generation from Multiple Action Labels. in B Williams, Y Chen & J Neville (eds), AAAI-23 Technical Tracks 1. Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023, vol. 37, AAAI press, pp. 1231-1239, 37th AAAI Conference on Artificial Intelligence, AAAI 2023, Washington, United States, 23/2/7. https://doi.org/10.1609/aaai.v37i1.25206

@inproceedings{cda959fe89fe4f9a838fa32c448e7d61,

title = "MultiAct: Long-Term 3D Human Motion Generation from Multiple Action Labels",

abstract = "We tackle the problem of generating long-term 3D human motion from multiple action labels. Two main previous approaches, such as action- and motion-conditioned methods, have limitations to solve this problem. The action-conditioned methods generate a sequence of motion from a single action. Hence, it cannot generate long-term motions composed of multiple actions and transitions between actions. Meanwhile, the motion-conditioned methods generate future motions from initial motion. The generated future motions only depend on the past, so they are not controllable by the user{\textquoteright}s desired actions. We present MultiAct, the first framework to generate long-term 3D human motion from multiple action labels. MultiAct takes account of both action and motion conditions with a unified recurrent generation system. It repetitively takes the previous motion and action label; then, it generates a smooth transition and the motion of the given action. As a result, MultiAct produces realistic long-term motion controlled by the given sequence of multiple action labels. Code is publicly available in https://github.com/TaeryungLee/MultiAct RELEASE.",

author = "Taeryung Lee and Gyeongsik Moon and Lee, \{Kyoung Mu\}",

note = "Publisher Copyright: Copyright {\textcopyright} 2023, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.; 37th AAAI Conference on Artificial Intelligence, AAAI 2023 ; Conference date: 07-02-2023 Through 14-02-2023",

year = "2023",

month = jun,

day = "27",

doi = "10.1609/aaai.v37i1.25206",

language = "English",

series = "Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023",

publisher = "AAAI press",

pages = "1231--1239",

editor = "Brian Williams and Yiling Chen and Jennifer Neville",

booktitle = "AAAI-23 Technical Tracks 1",

}

TY - GEN

T1 - MultiAct

T2 - 37th AAAI Conference on Artificial Intelligence, AAAI 2023

AU - Lee, Taeryung

AU - Moon, Gyeongsik

AU - Lee, Kyoung Mu

PY - 2023/6/27

Y1 - 2023/6/27

N2 - We tackle the problem of generating long-term 3D human motion from multiple action labels. Two main previous approaches, such as action- and motion-conditioned methods, have limitations to solve this problem. The action-conditioned methods generate a sequence of motion from a single action. Hence, it cannot generate long-term motions composed of multiple actions and transitions between actions. Meanwhile, the motion-conditioned methods generate future motions from initial motion. The generated future motions only depend on the past, so they are not controllable by the user’s desired actions. We present MultiAct, the first framework to generate long-term 3D human motion from multiple action labels. MultiAct takes account of both action and motion conditions with a unified recurrent generation system. It repetitively takes the previous motion and action label; then, it generates a smooth transition and the motion of the given action. As a result, MultiAct produces realistic long-term motion controlled by the given sequence of multiple action labels. Code is publicly available in https://github.com/TaeryungLee/MultiAct RELEASE.

AB - We tackle the problem of generating long-term 3D human motion from multiple action labels. Two main previous approaches, such as action- and motion-conditioned methods, have limitations to solve this problem. The action-conditioned methods generate a sequence of motion from a single action. Hence, it cannot generate long-term motions composed of multiple actions and transitions between actions. Meanwhile, the motion-conditioned methods generate future motions from initial motion. The generated future motions only depend on the past, so they are not controllable by the user’s desired actions. We present MultiAct, the first framework to generate long-term 3D human motion from multiple action labels. MultiAct takes account of both action and motion conditions with a unified recurrent generation system. It repetitively takes the previous motion and action label; then, it generates a smooth transition and the motion of the given action. As a result, MultiAct produces realistic long-term motion controlled by the given sequence of multiple action labels. Code is publicly available in https://github.com/TaeryungLee/MultiAct RELEASE.

UR - https://www.scopus.com/pages/publications/85160292345

U2 - 10.1609/aaai.v37i1.25206

DO - 10.1609/aaai.v37i1.25206

M3 - Conference contribution

AN - SCOPUS:85160292345

T3 - Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023

SP - 1231

EP - 1239

BT - AAAI-23 Technical Tracks 1

A2 - Williams, Brian

A2 - Chen, Yiling

A2 - Neville, Jennifer

PB - AAAI press

Y2 - 7 February 2023 through 14 February 2023

ER -

MultiAct: Long-Term 3D Human Motion Generation from Multiple Action Labels

Abstract

Publication series

Conference

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this