File Download
  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: Differentiable Learning-to-Group Channels via Groupable Convolutional Neural Networks

TitleDifferentiable Learning-to-Group Channels via Groupable Convolutional Neural Networks
Authors
Issue Date2019
PublisherInstitute of Electrical and Electronics Engineers. The Journal's web site is located at http://ieeexplore.ieee.org/xpl/conhome.jsp?punumber=1000149
Citation
Proceedings of IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea, 27 October - 2 November 2019, p. 3541-3550 How to Cite?
AbstractGroup convolution, which divides the channels of ConvNets into groups, has achieved impressive improvement over the regular convolution operation. However, existing models, eg ResNext, still suffers from the sub-optimal performance due to manually defining the number of groups as a constant over all of the layers. Toward addressing this issue, we present Groupable ConvNet (GroupNet) built by using a novel dynamic grouping convolution (DGConv) operation, which is able to learn the number of groups in an end-to-end manner. The proposed approach has several appealing benefits. (1) DGConv provides a unified convolution representation and covers many existing convolution operations such as regular dense convolution, group convolution, and depthwise convolution. (2) DGConv is a differentiable and flexible operation which learns to perform various convolutions from training data. (3) GroupNet trained with DGConv learns different number of groups for different convolution layers. Extensive experiments demonstrate that GroupNet outperforms its counterparts such as ResNet and ResNeXt in terms of accuracy and computational complexity. We also present introspection and reproducibility study, for the first time, showing the learning dynamics of training group numbers.
Persistent Identifierhttp://hdl.handle.net/10722/284158
ISSN

 

DC FieldValueLanguage
dc.contributor.authorZhang, Z-
dc.contributor.authorLi, J-
dc.contributor.authorShao, W-
dc.contributor.authorPeng, Z-
dc.contributor.authorZhang, R-
dc.contributor.authorWang, X-
dc.contributor.authorLuo, P-
dc.date.accessioned2020-07-20T05:56:33Z-
dc.date.available2020-07-20T05:56:33Z-
dc.date.issued2019-
dc.identifier.citationProceedings of IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea, 27 October - 2 November 2019, p. 3541-3550-
dc.identifier.issn1550-5499-
dc.identifier.urihttp://hdl.handle.net/10722/284158-
dc.description.abstractGroup convolution, which divides the channels of ConvNets into groups, has achieved impressive improvement over the regular convolution operation. However, existing models, eg ResNext, still suffers from the sub-optimal performance due to manually defining the number of groups as a constant over all of the layers. Toward addressing this issue, we present Groupable ConvNet (GroupNet) built by using a novel dynamic grouping convolution (DGConv) operation, which is able to learn the number of groups in an end-to-end manner. The proposed approach has several appealing benefits. (1) DGConv provides a unified convolution representation and covers many existing convolution operations such as regular dense convolution, group convolution, and depthwise convolution. (2) DGConv is a differentiable and flexible operation which learns to perform various convolutions from training data. (3) GroupNet trained with DGConv learns different number of groups for different convolution layers. Extensive experiments demonstrate that GroupNet outperforms its counterparts such as ResNet and ResNeXt in terms of accuracy and computational complexity. We also present introspection and reproducibility study, for the first time, showing the learning dynamics of training group numbers.-
dc.languageeng-
dc.publisherInstitute of Electrical and Electronics Engineers. The Journal's web site is located at http://ieeexplore.ieee.org/xpl/conhome.jsp?punumber=1000149-
dc.relation.ispartofIEEE International Conference on Computer Vision (ICCV) Proceedings-
dc.rightsIEEE International Conference on Computer Vision (ICCV) Proceedings. Copyright © Institute of Electrical and Electronics Engineers.-
dc.rights©2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.-
dc.titleDifferentiable Learning-to-Group Channels via Groupable Convolutional Neural Networks-
dc.typeConference_Paper-
dc.identifier.emailLuo, P: pluo@hku.hk-
dc.identifier.authorityLuo, P=rp02575-
dc.description.naturepostprint-
dc.identifier.doi10.1109/ICCV.2019.00364-
dc.identifier.hkuros311018-
dc.identifier.spage3541-
dc.identifier.epage3550-
dc.publisher.placeUnited States-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats