Integrating Generative and Contrastive Approaches for Human Action Recognition

This study introduces a novel approach to unsupervised skeleton-based human action recognition by integrating generative and contrastive learning methods. We propose a decomposition of representations, allowing for the preservation of detailed motion information for the generative learning objective...

Full description

Saved in:
Bibliographic Details
Main Authors: Pablo Cervantes, Yusuke Sekikawa, Ikuro Sato, Koichi Shinoda
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11020639/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850130463605129216
author Pablo Cervantes
Yusuke Sekikawa
Ikuro Sato
Koichi Shinoda
author_facet Pablo Cervantes
Yusuke Sekikawa
Ikuro Sato
Koichi Shinoda
author_sort Pablo Cervantes
collection DOAJ
description This study introduces a novel approach to unsupervised skeleton-based human action recognition by integrating generative and contrastive learning methods. We propose a decomposition of representations, allowing for the preservation of detailed motion information for the generative learning objective while also extracting action features for the contrastive learning objective. By swapping contrastive representations between positive pairs (coining the name SwapCLR), we ensure that the generative and contrastive representations are complementary and both objectives contribute to learning a strong representation for downstream tasks like action recognition. Additionally, we address the challenge of noisy data in skeleton-based action recognition with a new saturating reconstruction loss, significantly reducing the impact of noise common to key-point detections. Our method demonstrates state-of-the-art performance in unsupervised action recognition on the NTU and PKU-MMD datasets, while also enabling generative downstream tasks such as motion in-painting and motion generation. Overall, these experimental results confirm the method’s effectiveness and suggest its applicability to a variety of action analysis tasks.
format Article
id doaj-art-9fd955e4ce9e42b585ca5ca393e1e17e
institution OA Journals
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-9fd955e4ce9e42b585ca5ca393e1e17e2025-08-20T02:32:41ZengIEEEIEEE Access2169-35362025-01-011310009510010410.1109/ACCESS.2025.357570711020639Integrating Generative and Contrastive Approaches for Human Action RecognitionPablo Cervantes0https://orcid.org/0000-0002-5256-9317Yusuke Sekikawa1https://orcid.org/0000-0003-1111-5949Ikuro Sato2https://orcid.org/0000-0001-5234-3177Koichi Shinoda3https://orcid.org/0000-0003-1095-3203Institute of Science Tokyo (formerly Tokyo Institute of Technology), Tokyo, JapanDenso IT Laboratory, Minato City, JapanInstitute of Science Tokyo (formerly Tokyo Institute of Technology), Tokyo, JapanInstitute of Science Tokyo (formerly Tokyo Institute of Technology), Tokyo, JapanThis study introduces a novel approach to unsupervised skeleton-based human action recognition by integrating generative and contrastive learning methods. We propose a decomposition of representations, allowing for the preservation of detailed motion information for the generative learning objective while also extracting action features for the contrastive learning objective. By swapping contrastive representations between positive pairs (coining the name SwapCLR), we ensure that the generative and contrastive representations are complementary and both objectives contribute to learning a strong representation for downstream tasks like action recognition. Additionally, we address the challenge of noisy data in skeleton-based action recognition with a new saturating reconstruction loss, significantly reducing the impact of noise common to key-point detections. Our method demonstrates state-of-the-art performance in unsupervised action recognition on the NTU and PKU-MMD datasets, while also enabling generative downstream tasks such as motion in-painting and motion generation. Overall, these experimental results confirm the method’s effectiveness and suggest its applicability to a variety of action analysis tasks.https://ieeexplore.ieee.org/document/11020639/Generative and contrastiverepresentation learningunsupervised 3D action recognition
spellingShingle Pablo Cervantes
Yusuke Sekikawa
Ikuro Sato
Koichi Shinoda
Integrating Generative and Contrastive Approaches for Human Action Recognition
IEEE Access
Generative and contrastive
representation learning
unsupervised 3D action recognition
title Integrating Generative and Contrastive Approaches for Human Action Recognition
title_full Integrating Generative and Contrastive Approaches for Human Action Recognition
title_fullStr Integrating Generative and Contrastive Approaches for Human Action Recognition
title_full_unstemmed Integrating Generative and Contrastive Approaches for Human Action Recognition
title_short Integrating Generative and Contrastive Approaches for Human Action Recognition
title_sort integrating generative and contrastive approaches for human action recognition
topic Generative and contrastive
representation learning
unsupervised 3D action recognition
url https://ieeexplore.ieee.org/document/11020639/
work_keys_str_mv AT pablocervantes integratinggenerativeandcontrastiveapproachesforhumanactionrecognition
AT yusukesekikawa integratinggenerativeandcontrastiveapproachesforhumanactionrecognition
AT ikurosato integratinggenerativeandcontrastiveapproachesforhumanactionrecognition
AT koichishinoda integratinggenerativeandcontrastiveapproachesforhumanactionrecognition