File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: DeciWatch: A Simple Baseline for 10 × Efficient 2D and 3D Pose Estimation

TitleDeciWatch: A Simple Baseline for 10 × Efficient 2D and 3D Pose Estimation
Authors
KeywordsEfficiency
Human Pose Estimation
Video Analysis
Issue Date2022
Citation
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2022, v. 13665 LNCS, p. 607-624 How to Cite?
AbstractThis paper proposes a simple baseline framework for video-based 2D/3D human pose estimation that can achieve 10 × efficiency improvement over existing works without any performance degradation, named DeciWatch. Unlike current solutions that estimate each frame in a video, DeciWatch introduces a simple yet effective sample-denoise-recover framework that only watches sparsely sampled frames, taking advantage of the continuity of human motions and the lightweight pose representation. Specifically, DeciWatch uniformly samples less than 10 % video frames for detailed estimation, denoises the estimated 2D/3D poses with an efficient Transformer architecture, and then accurately recovers the rest of the frames using another Transformer-based network. Comprehensive experimental results on three video-based human pose estimation, body mesh recovery tasks and efficient labeling in videos with four datasets validate the efficiency and effectiveness of DeciWatch. Code is available at https://github.com/cure-lab/DeciWatch.
Persistent Identifierhttp://hdl.handle.net/10722/352336
ISSN
2023 SCImago Journal Rankings: 0.606

 

DC FieldValueLanguage
dc.contributor.authorZeng, Ailing-
dc.contributor.authorJu, Xuan-
dc.contributor.authorYang, Lei-
dc.contributor.authorGao, Ruiyuan-
dc.contributor.authorZhu, Xizhou-
dc.contributor.authorDai, Bo-
dc.contributor.authorXu, Qiang-
dc.date.accessioned2024-12-16T03:58:20Z-
dc.date.available2024-12-16T03:58:20Z-
dc.date.issued2022-
dc.identifier.citationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2022, v. 13665 LNCS, p. 607-624-
dc.identifier.issn0302-9743-
dc.identifier.urihttp://hdl.handle.net/10722/352336-
dc.description.abstractThis paper proposes a simple baseline framework for video-based 2D/3D human pose estimation that can achieve 10 × efficiency improvement over existing works without any performance degradation, named DeciWatch. Unlike current solutions that estimate each frame in a video, DeciWatch introduces a simple yet effective sample-denoise-recover framework that only watches sparsely sampled frames, taking advantage of the continuity of human motions and the lightweight pose representation. Specifically, DeciWatch uniformly samples less than 10 % video frames for detailed estimation, denoises the estimated 2D/3D poses with an efficient Transformer architecture, and then accurately recovers the rest of the frames using another Transformer-based network. Comprehensive experimental results on three video-based human pose estimation, body mesh recovery tasks and efficient labeling in videos with four datasets validate the efficiency and effectiveness of DeciWatch. Code is available at https://github.com/cure-lab/DeciWatch.-
dc.languageeng-
dc.relation.ispartofLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)-
dc.subjectEfficiency-
dc.subjectHuman Pose Estimation-
dc.subjectVideo Analysis-
dc.titleDeciWatch: A Simple Baseline for 10 × Efficient 2D and 3D Pose Estimation-
dc.typeConference_Paper-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1007/978-3-031-20065-6_35-
dc.identifier.scopuseid_2-s2.0-85144490710-
dc.identifier.volume13665 LNCS-
dc.identifier.spage607-
dc.identifier.epage624-
dc.identifier.eissn1611-3349-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats