Show simple item record

dc.contributor.author
Zhao, Mengyi
dc.contributor.author
Dai, Shuling
dc.contributor.author
Zhu, Yanjun
dc.contributor.author
Tang, Hao
dc.contributor.author
Xie, Pan
dc.contributor.author
Li, Yue
dc.contributor.author
Liu, Chunlei
dc.contributor.author
Zhang, Baochang
dc.date.accessioned
2022-08-09T13:41:25Z
dc.date.available
2022-07-10T03:30:57Z
dc.date.available
2022-08-09T13:41:25Z
dc.date.issued
2022-08-28
dc.identifier.issn
0925-2312
dc.identifier.issn
1872-8286
dc.identifier.other
10.1016/j.neucom.2022.06.070
en_US
dc.identifier.uri
http://hdl.handle.net/20.500.11850/557164
dc.description.abstract
Skeleton-based action recognition is an essential yet challenging visual task, whose accuracy has been remarkably improved due to the successful application of graph convolutional networks (GCNs). However, high computation cost and memory usage hinder their deployment on resource-constrained environment. To deal with the issue, in this paper, we introduce two novel progressive binary graph convolutional network for skeleton-based action recognition PB-GCN and PB-GCN*, which can obtain significant speed-up and memory saving. In PB-GCN, the filters are binarized, and in PB-GCN*, both filters and activations are binary. Specifically, we propose a progressive optimization, i.e., employing ternary models as the initialization of binary GCNs (BGCN) to improve the representational capability of binary models. Moreover, the center loss is exploited to improve the training procedure for better performance. Experimental results on two public benchmarks (i.e., Skeleton-Kinetics and NTU RGB + D) demonstrate that the accuracy of the proposed PB-GCN and PB-GCN* are comparable to their full-precision counterparts and outperforms the state-of-the-art methods, such as BWN, XNOR-Net, and Bi-Real Net.
en_US
dc.language.iso
en
en_US
dc.publisher
Elsevier
en_US
dc.subject
Binary neural network
en_US
dc.subject
Center loss
en_US
dc.subject
Progressive optimization
en_US
dc.subject
Skeleton-based action recognition
en_US
dc.subject
Spatial-temporal graph convolutional networks
en_US
dc.title
PB-GCN: Progressive binary graph convolutional networks for skeleton-based action recognition
en_US
dc.type
Journal Article
dc.date.published
2022-06-23
ethz.journal.title
Neurocomputing
ethz.journal.volume
501
en_US
ethz.journal.abbreviated
Neurocomputing
ethz.pages.start
640
en_US
ethz.pages.end
649
en_US
ethz.identifier.wos
ethz.identifier.scopus
ethz.publication.place
Amsterdam
en_US
ethz.publication.status
published
en_US
ethz.date.deposited
2022-07-10T03:30:59Z
ethz.source
SCOPUS
ethz.eth
yes
en_US
ethz.availability
Metadata only
en_US
ethz.rosetta.installDate
2022-08-09T13:41:33Z
ethz.rosetta.lastUpdated
2022-08-09T13:41:33Z
ethz.rosetta.versionExported
true
ethz.COinS
ctx_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.atitle=PB-GCN:%20Progressive%20binary%20graph%20convolutional%20networks%20for%20skeleton-based%20action%20recognition&rft.jtitle=Neurocomputing&rft.date=2022-08-28&rft.volume=501&rft.spage=640&rft.epage=649&rft.issn=0925-2312&1872-8286&rft.au=Zhao,%20Mengyi&Dai,%20Shuling&Zhu,%20Yanjun&Tang,%20Hao&Xie,%20Pan&rft.genre=article&rft_id=info:doi/10.1016/j.neucom.2022.06.070&
 Search print copy at ETH Library

Files in this item

FilesSizeFormatOpen in viewer

There are no files associated with this item.

Publication type

Show simple item record