RB-Net: Training Highly Accurate and Efficient Binary Neural Networks With Reshaped Point-Wise Convolution and Balanced Activation

In this paper, we find that the conventional convolution operation becomes the bottleneck for extremely efficient binary neural networks (BNNs). To address this issue, we open up a new direction by introducing a reshaped point-wise convolution (RPC) to replace the conventional one to build BNNs. Spe...

Full description

Saved in:

Bibliographic Details
Published in:	IEEE transactions on circuits and systems for video technology Vol. 32; no. 9; pp. 6414 - 6424
Main Authors:	Liu, Chunlei, Ding, Wenrui, Chen, Peng, Zhuang, Bohan, Wang, Yufeng, Zhao, Yang, Zhang, Baochang, Han, Yuqi
Format:	Journal Article
Language:	English
Published:	New York IEEE 01.09.2022 The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:	Accuracy balanced activation Binary neural network Complexity Computational efficiency Computing costs Convolution Kernel Neural networks object classification Remote procedure calls reshaped point-wise convolution Spatial data Spatial databases Training
ISSN:	1051-8215, 1558-2205
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Abstract	In this paper, we find that the conventional convolution operation becomes the bottleneck for extremely efficient binary neural networks (BNNs). To address this issue, we open up a new direction by introducing a reshaped point-wise convolution (RPC) to replace the conventional one to build BNNs. Specifically, we conduct a point-wise convolution after rearranging the spatial information into depth, with which at least <inline-formula> <tex-math notation="LaTeX">2.25\times </tex-math></inline-formula> computation reduction can be achieved. Such an efficient RPC allows us to explore more powerful representational capacity of BNNs under a given computation complexity budget. Moreover, we propose to use a balanced activation (BA) to adjust the distribution of the scaled activations after binarization, which enables significant performance improvement of BNNs. After integrating RPC and BA, the proposed network, dubbed as RB-Net, strikes a good trade-off between accuracy and efficiency, achieving superior performance with lower computational cost against the state-of-the-art BNN methods. Specifically, our RB-Net achieves 66.8% Top-1 accuracy with ResNet-18 backbone on ImageNet, exceeding the state-of-the-art Real-to-Binary Net (65.4%) by 1.4% while achieving more than <inline-formula> <tex-math notation="LaTeX">3\times </tex-math></inline-formula> reduction (52M vs. 165M) in computational complexity.
AbstractList	In this paper, we find that the conventional convolution operation becomes the bottleneck for extremely efficient binary neural networks (BNNs). To address this issue, we open up a new direction by introducing a reshaped point-wise convolution (RPC) to replace the conventional one to build BNNs. Specifically, we conduct a point-wise convolution after rearranging the spatial information into depth, with which at least <inline-formula> <tex-math notation="LaTeX">2.25\times </tex-math></inline-formula> computation reduction can be achieved. Such an efficient RPC allows us to explore more powerful representational capacity of BNNs under a given computation complexity budget. Moreover, we propose to use a balanced activation (BA) to adjust the distribution of the scaled activations after binarization, which enables significant performance improvement of BNNs. After integrating RPC and BA, the proposed network, dubbed as RB-Net, strikes a good trade-off between accuracy and efficiency, achieving superior performance with lower computational cost against the state-of-the-art BNN methods. Specifically, our RB-Net achieves 66.8% Top-1 accuracy with ResNet-18 backbone on ImageNet, exceeding the state-of-the-art Real-to-Binary Net (65.4%) by 1.4% while achieving more than <inline-formula> <tex-math notation="LaTeX">3\times </tex-math></inline-formula> reduction (52M vs. 165M) in computational complexity. In this paper, we find that the conventional convolution operation becomes the bottleneck for extremely efficient binary neural networks (BNNs). To address this issue, we open up a new direction by introducing a reshaped point-wise convolution (RPC) to replace the conventional one to build BNNs. Specifically, we conduct a point-wise convolution after rearranging the spatial information into depth, with which at least [Formula Omitted] computation reduction can be achieved. Such an efficient RPC allows us to explore more powerful representational capacity of BNNs under a given computation complexity budget. Moreover, we propose to use a balanced activation (BA) to adjust the distribution of the scaled activations after binarization, which enables significant performance improvement of BNNs. After integrating RPC and BA, the proposed network, dubbed as RB-Net, strikes a good trade-off between accuracy and efficiency, achieving superior performance with lower computational cost against the state-of-the-art BNN methods. Specifically, our RB-Net achieves 66.8% Top-1 accuracy with ResNet-18 backbone on ImageNet, exceeding the state-of-the-art Real-to-Binary Net (65.4%) by 1.4% while achieving more than [Formula Omitted] reduction (52M vs. 165M) in computational complexity.
Author	Ding, Wenrui Wang, Yufeng Zhang, Baochang Liu, Chunlei Chen, Peng Zhao, Yang Zhuang, Bohan Han, Yuqi
Author_xml	– sequence: 1 givenname: Chunlei surname: Liu fullname: Liu, Chunlei organization: School of Electrical and Information Engineering, Beihang University, Beijing, China – sequence: 2 givenname: Wenrui orcidid: 0000-0001-5490-4724 surname: Ding fullname: Ding, Wenrui organization: Unmanned System Research Institute, Beihang University, Beijing, China – sequence: 3 givenname: Peng surname: Chen fullname: Chen, Peng organization: Australian Institute for Machine Learning (AIML), The University of Adelaide, Adelaide, SA, Australia – sequence: 4 givenname: Bohan surname: Zhuang fullname: Zhuang, Bohan organization: Department of Information Tehnology, Monash University Clayton Campus, Melbourne, VIC, Australia – sequence: 5 givenname: Yufeng orcidid: 0000-0001-8713-3153 surname: Wang fullname: Wang, Yufeng organization: Australian Institute for Machine Learning (AIML), The University of Adelaide, Adelaide, SA, Australia – sequence: 6 givenname: Yang orcidid: 0000-0001-5252-658X surname: Zhao fullname: Zhao, Yang organization: Australian Institute for Machine Learning (AIML), The University of Adelaide, Adelaide, SA, Australia – sequence: 7 givenname: Baochang orcidid: 0000-0001-7396-6218 surname: Zhang fullname: Zhang, Baochang organization: Institute of Artificial Intelligence, Beihang University, Beijing, China – sequence: 8 givenname: Yuqi orcidid: 0000-0001-7905-0163 surname: Han fullname: Han, Yuqi email: yuqi_han@tsinghua.edu.cn organization: Department of Computer Science and Technology, Tsinghua University, Beijing, China
BookMark	eNp9kEFPGzEQha0KpALtH6AXSz1vsL2x1-4tiWiphCiCFI6r2c2YmG7tYDsgrvxynAT10AOnGc2890bzHZI9HzwScszZiHNmTuaz65v5SDAhRjVXSrP6AzngUupKCCb3Ss8kr7Tg8iM5TOmeMT7W4-aAvFxNqwvM3-g8gvPO39Ezd7ccnumk79cRMlLwC3pqresd-kynzkN8phdYlkMp-SnEP4neurykV5iWsMIFvQzO5-rWJaSz4B_DsM4u-G3SFAbwfdFM-uweYTP_RPYtDAk_v9Uj8vv76Xx2Vp3_-vFzNjmvemFkrhbQWSNFo2XNNDd2gQqtMKCxAWu47iwHQGEBpFJKdkZ1VjX1uJNjBqY29RH5ustdxfCwxpTb-7COvpxsRcMMlyVVF5XYqfoYUopo21V0f8vPLWfthnW7Zd1uWLdvrItJ_2fqXd4-lwvW4X3rl53VIeK_W6aRijFWvwLXIpAK
CODEN	ITCTEM
CitedBy_id	crossref_primary_10_1109_TCSVT_2023_3301962 crossref_primary_10_1109_TCSVT_2023_3288572 crossref_primary_10_3390_rs14215347 crossref_primary_10_1109_TCSVT_2023_3279281 crossref_primary_10_1109_TCSVT_2023_3286072 crossref_primary_10_1007_s10489_024_06039_z crossref_primary_10_1109_TGRS_2025_3540784 crossref_primary_10_26599_BDMA_2024_9020087 crossref_primary_10_1007_s11554_023_01366_9 crossref_primary_10_1109_TNSRE_2023_3321414 crossref_primary_10_1109_ACCESS_2023_3258360 crossref_primary_10_1109_TCSVT_2024_3399311 crossref_primary_10_3390_s23042250 crossref_primary_10_3390_s24248190 crossref_primary_10_1109_TCSVT_2022_3216389 crossref_primary_10_1109_JSTARS_2022_3216564 crossref_primary_10_3389_fnins_2022_1080521 crossref_primary_10_3390_electronics13020266 crossref_primary_10_3390_s24010181 crossref_primary_10_1109_LGRS_2023_3308582
Cites_doi	10.1109/ICCV.2017.155 10.1109/CVPR.2019.00050 10.1109/CVPR.2019.00448 10.1109/CVPR.2018.00452 10.1145/3343031.3350534 10.1609/aaai.v33i01.33018344 10.1109/CVPR.2018.00826 10.1007/978-3-030-01267-0_44 10.1109/CVPR.2016.90 10.1109/CVPR.2018.00286 10.1109/TCSVT.2019.2911674 10.1109/CVPR.2018.00716 10.1109/ICCV.2019.00501 10.1109/ICASSP.2018.8461456 10.1007/978-3-030-01237-3_23 10.1109/TCSVT.2019.2896095 10.1007/978-3-030-58568-6_9 10.1109/TCSI.2019.2958937 10.1109/ICCV.2019.00495 10.1007/978-3-319-46493-0_32 10.1109/CVPR.2019.01167 10.1109/TCSVT.2020.3040367 10.1109/TCSVT.2020.2996231 10.1109/TCSVT.2021.3095970 10.1109/CVPR.2019.00066 10.1609/aaai.v31i1.10862 10.1007/s11263-015-0816-y 10.1109/CVPR.2017.574 10.1109/CVPR.2018.00474
ContentType	Journal Article
Copyright	Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2022
Copyright_xml	– notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2022
DBID	97E RIA RIE AAYXX CITATION 7SC 7SP 8FD JQ2 L7M L~C L~D
DOI	10.1109/TCSVT.2022.3166803
DatabaseName	IEEE Xplore (IEEE) IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE Electronic Library (IEL) CrossRef Computer and Information Systems Abstracts Electronics & Communications Abstracts Technology Research Database ProQuest Computer Science Collection Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Academic Computer and Information Systems Abstracts Professional
DatabaseTitle	CrossRef Technology Research Database Computer and Information Systems Abstracts – Academic Electronics & Communications Abstracts ProQuest Computer Science Collection Computer and Information Systems Abstracts Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Professional
DatabaseTitleList	Technology Research Database
Database_xml	– sequence: 1 dbid: RIE name: IEEE Electronic Library (IEL) url: https://ieeexplore.ieee.org/ sourceTypes: Publisher
DeliveryMethod	fulltext_linktorsrc
Discipline	Engineering
EISSN	1558-2205
EndPage	6424
ExternalDocumentID	10_1109_TCSVT_2022_3166803 9756000
Genre	orig-research
GrantInformation_xml	– fundername: China Postdoctoral Science Foundation grantid: 2021TQ0177 funderid: 10.13039/501100002858
GroupedDBID	-~X 0R~ 29I 4.4 5GY 5VS 6IK 97E AAJGR AARMG AASAJ AAWTH ABAZT ABQJQ ABVLG ACGFO ACGFS ACIWK AENEX AETIX AGQYO AGSQL AHBIQ AI. AIBXA AKJIK AKQYR ALLEH ALMA_UNASSIGNED_HOLDINGS ASUFR ATWAV BEFXN BFFAM BGNUA BKEBE BPEOZ CS3 DU5 EBS EJD HZ~ H~9 ICLAB IFIPE IFJZH IPLJI JAVBF LAI M43 O9- OCL P2P RIA RIE RNS RXW TAE TN5 VH1 AAYXX CITATION 7SC 7SP 8FD JQ2 L7M L~C L~D
ID	FETCH-LOGICAL-c295t-dabf95278530819fde6ef29a8e7af918bf1aae2faa56665b96bf6734b540a9393
IEDL.DBID	RIE
ISICitedReferencesCount	25
ISICitedReferencesURI	http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000849300000056&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN	1051-8215
IngestDate	Sun Nov 30 04:06:40 EST 2025 Sat Nov 29 01:44:18 EST 2025 Tue Nov 18 21:31:34 EST 2025 Wed Aug 27 02:14:21 EDT 2025
IsPeerReviewed	true
IsScholarly	true
Issue	9
Language	English
License	https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html https://doi.org/10.15223/policy-029 https://doi.org/10.15223/policy-037
LinkModel	DirectLink
MergedId	FETCHMERGED-LOGICAL-c295t-dabf95278530819fde6ef29a8e7af918bf1aae2faa56665b96bf6734b540a9393
Notes	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14
ORCID	0000-0001-7396-6218 0000-0001-5252-658X 0000-0001-7905-0163 0000-0001-8713-3153 0000-0001-5490-4724
PQID	2709158198
PQPubID	85433
PageCount	11
ParticipantIDs	crossref_primary_10_1109_TCSVT_2022_3166803 ieee_primary_9756000 proquest_journals_2709158198 crossref_citationtrail_10_1109_TCSVT_2022_3166803
PublicationCentury	2000
PublicationDate	2022-09-01
PublicationDateYYYYMMDD	2022-09-01
PublicationDate_xml	– month: 09 year: 2022 text: 2022-09-01 day: 01
PublicationDecade	2020
PublicationPlace	New York
PublicationPlace_xml	– name: New York
PublicationTitle	IEEE transactions on circuits and systems for video technology
PublicationTitleAbbrev	TCSVT
PublicationYear	2022
Publisher	IEEE The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Publisher_xml	– name: IEEE – name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
References	ref13 ref12 ref15 ref14 ref11 ref16 Bethge (ref18) 2020 ref19 Zhou (ref23) 2017 Darabi (ref41) 2018 Courbariaux (ref8) 2016 Howard (ref21) 2017 ref51 ref48 Bulat (ref10) 2019 ref47 Zhou (ref7) 2016 ref43 Martinez (ref17) Ridnik (ref45) 2020 ref9 Mishra (ref26) 2017 ref4 ref3 ref6 ref5 Helwegen (ref35) Qin (ref22) 2019 ref34 ref36 ref31 ref30 ref33 ref32 ref2 ref1 Choi (ref25) 2018 ref39 Krizhevsky (ref42); 25 Yu (ref44) Hinton (ref40) 2012 Wu (ref46) ref24 ref20 Bulat (ref37) 2019 ref28 ref27 ref29 Kingma (ref49) 2014 Hou (ref38) 2016 Simonyan (ref50) 2014
References_xml	– year: 2019 ident: ref22 article-title: Forward and backward information retention for accurate binary neural networks publication-title: arXiv:1909.10788 – ident: ref32 doi: 10.1109/ICCV.2017.155 – ident: ref14 doi: 10.1109/CVPR.2019.00050 – year: 2016 ident: ref7 article-title: DoReFa-Net: Training low bitwidth convolutional neural networks with low bitwidth gradients publication-title: arXiv:1606.06160 – ident: ref30 doi: 10.1109/CVPR.2019.00448 – ident: ref29 doi: 10.1109/CVPR.2018.00452 – ident: ref31 doi: 10.1145/3343031.3350534 – volume-title: ICLR ident: ref17 article-title: Training binary neural networks with real-to-binary convolutions – year: 2020 ident: ref18 article-title: MeliusNet: Can binary neural networks achieve MobileNet-level accuracy? publication-title: arXiv:2001.05936 – year: 2019 ident: ref37 article-title: XNOR-Net++: Improved binary neural networks publication-title: arXiv:1909.13863 – ident: ref36 doi: 10.1609/aaai.v33i01.33018344 – ident: ref16 doi: 10.1109/CVPR.2018.00826 – year: 2019 ident: ref10 article-title: Matrix and tensor decompositions for training binary neural networks publication-title: arXiv:1904.07852 – ident: ref13 doi: 10.1007/978-3-030-01267-0_44 – year: 2018 ident: ref41 article-title: Regularized binary network training publication-title: arXiv:1812.11800 – ident: ref48 doi: 10.1109/CVPR.2016.90 – year: 2017 ident: ref23 article-title: Incremental network quantization: Towards lossless CNNs with low-precision weights publication-title: arXiv:1702.03044 – year: 2020 ident: ref45 article-title: TResNet: High performance GPU-dedicated architecture publication-title: arXiv:2003.13630 – ident: ref34 doi: 10.1109/CVPR.2018.00286 – ident: ref5 doi: 10.1109/TCSVT.2019.2911674 – year: 2016 ident: ref8 article-title: Binarized neural networks: Training deep neural networks with weights and activations constrained to +1 or −1 publication-title: arXiv:1602.02830 – ident: ref33 doi: 10.1109/CVPR.2018.00716 – start-page: 1 volume-title: Proc. ICLR ident: ref44 article-title: Multi-scale context aggregation by dilated convolutions – year: 2014 ident: ref49 article-title: Adam: A method for stochastic optimization publication-title: arXiv:1412.6980 – ident: ref39 doi: 10.1109/ICCV.2019.00501 – ident: ref11 doi: 10.1109/ICASSP.2018.8461456 – ident: ref24 doi: 10.1007/978-3-030-01237-3_23 – ident: ref2 doi: 10.1109/TCSVT.2019.2896095 – ident: ref19 doi: 10.1007/978-3-030-58568-6_9 – volume: 25 start-page: 1 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref42 article-title: ImageNet classification with deep convolutional neural networks – ident: ref1 doi: 10.1109/TCSI.2019.2958937 – ident: ref27 doi: 10.1109/ICCV.2019.00495 – year: 2016 ident: ref38 article-title: Loss-aware binarization of deep networks publication-title: arXiv:1611.01600 – ident: ref12 doi: 10.1007/978-3-319-46493-0_32 – year: 2012 ident: ref40 article-title: Coursera lectures: Neural networks for machine learning – year: 2014 ident: ref50 article-title: Very deep convolutional networks for large-scale image recognition publication-title: arXiv:1409.1556 – ident: ref15 doi: 10.1109/CVPR.2019.01167 – year: 2017 ident: ref26 article-title: WRPN: Wide reduced-precision networks publication-title: arXiv:1709.01134 – year: 2017 ident: ref21 article-title: MobileNets: Efficient convolutional neural networks for mobile vision applications publication-title: arXiv:1704.04861 – start-page: 1194 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref46 article-title: Convolution with even-sized kernels and symmetric padding – ident: ref3 doi: 10.1109/TCSVT.2020.3040367 – ident: ref4 doi: 10.1109/TCSVT.2020.2996231 – ident: ref6 doi: 10.1109/TCSVT.2021.3095970 – ident: ref51 doi: 10.1109/CVPR.2019.00066 – ident: ref20 doi: 10.1007/978-3-030-58568-6_9 – ident: ref9 doi: 10.1609/aaai.v31i1.10862 – ident: ref47 doi: 10.1007/s11263-015-0816-y – ident: ref28 doi: 10.1109/CVPR.2017.574 – ident: ref43 doi: 10.1109/CVPR.2018.00474 – start-page: 7531 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref35 article-title: Latent weights do not exist: Rethinking binarized neural network optimization – year: 2018 ident: ref25 article-title: PACT: Parameterized clipping activation for quantized neural networks publication-title: arXiv:1805.06085
SSID	ssj0014847
Score	2.5229845
Snippet	In this paper, we find that the conventional convolution operation becomes the bottleneck for extremely efficient binary neural networks (BNNs). To address...
SourceID	proquest crossref ieee
SourceType	Aggregation Database Enrichment Source Index Database Publisher
StartPage	6414
SubjectTerms	Accuracy balanced activation Binary neural network Complexity Computational efficiency Computing costs Convolution Kernel Neural networks object classification Remote procedure calls reshaped point-wise convolution Spatial data Spatial databases Training
Title	RB-Net: Training Highly Accurate and Efficient Binary Neural Networks With Reshaped Point-Wise Convolution and Balanced Activation
URI	https://ieeexplore.ieee.org/document/9756000 https://www.proquest.com/docview/2709158198
Volume	32
WOSCitedRecordID	wos000849300000056&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText	1
inHoldings	1
isFullTextHit
isPrint
journalDatabaseRights	– providerCode: PRVIEE databaseName: IEEE Electronic Library (IEL) customDbUrl: eissn: 1558-2205 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0014847 issn: 1051-8215 databaseCode: RIE dateStart: 19910101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE
link	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1NTxsxELVo1EN7AApUBNLKh96KIev9sM0tiYJ6qCJEA-S28qeIhDYo2UTiyi_H4_VGSK0qcdvDeLSr5915s555g9APECWSHPTtcl6QzKYFEZlQRHky0ZdAYF1A-jebTPhsJq530Nm2F8ZaG4rP7DlchrN8s9Br-FV2IRjEZ5-gf2CMNb1a2xODjIdhYp4uJIT7ONY2yPTFxXT0527qU0FKfYZaFLwdkBWDUJiq8tenOMSXq7333dk-2o08Eg8a4L-gHVsdoM9v1AUP0cvNkExsfYmncQwEhqKOx2c80HoNChFYVgaPg4aE946HoTUXg1yH9zxp6sNX-H5eP-AbOBl6sgZfL-ZVTe7nK4tHi2oTN27wNIQqSe1tBrqdmXaEbq_G09EvEkcuEE1FXhMjlRM5ZT6IA1dwxhbWUSG5ZdKJhCuXSGmpk9LTwCJXolCuYGmmPPGTIhXpV9SpFpU9Rpjr1CllsgQkcLhgwvhkjGeUGs-gMmG6KGkxKHXUI4exGI9lyEv6ogy4lYBbGXHrop_bNU-NGsd_rQ8Bqa1lBKmLei3UZXxhVyVlnjjl_pH5yb9XnaJP4LspL-uhTr1c22_oo97U89Xye9iLr5Ji2xs
linkProvider	IEEE
linkToHtml	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1NT9wwEB0hQGo5AC2tWEqLD721LhvHSezedlcgULcRomnhFvlTXQllEZtF4sovx3acVSWqSr3l4DiJnpN5E8-8B_DRixIJ5vXtMpZjatIcc8ollo5MDIUnsDYgPS3Kkl1f84s1-LzqhTHGhOIz88Ufhr18PVdL_6vsmBc-PrsEfSOjlCRdt9Zqz4CyYCfmCEOCmYtkfYvMkB9Xkx-_KpcMEuJy1DxnvUVWDEPBV-XZxzhEmNOd_7u3XdiOTBKNOuhfwZppXsPWH_qCe_B4Ocalab-iKhpBIF_WcfOARkotvUYEEo1GJ0FFws2OxqE5F3nBDjdz2VWIL9DVrP2NLv3e0K3R6GI-a1p8NVsYNJk393HphpnGvk5SuTEj1bumvYGfpyfV5AxH0wWsCM9arIW0PCOFC-OeLVhtcmMJF8wUwvKESZsIYYgVwhHBPJM8lzYvUiod9RM85elbWG_mjdkHxFRqpdQ08SI4jBdcu3SMUUK041CU6wEkPQa1iork3hjjpg6ZyZDXAbfa41ZH3AbwaXXObafH8c_Rex6p1cgI0gAOe6jr-MoualI46pS5R2YHfz_rCF6cVd-n9fS8_PYOXvrrdMVmh7De3i3Ne9hU9-1scfchrMsnHTreYg
openUrl	ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=RB-Net%3A+Training+Highly+Accurate+and+Efficient+Binary+Neural+Networks+With+Reshaped+Point-Wise+Convolution+and+Balanced+Activation&rft.jtitle=IEEE+transactions+on+circuits+and+systems+for+video+technology&rft.au=Liu%2C+Chunlei&rft.au=Ding%2C+Wenrui&rft.au=Chen%2C+Peng&rft.au=Zhuang%2C+Bohan&rft.date=2022-09-01&rft.pub=IEEE&rft.issn=1051-8215&rft.volume=32&rft.issue=9&rft.spage=6414&rft.epage=6424&rft_id=info:doi/10.1109%2FTCSVT.2022.3166803&rft.externalDocID=9756000
thumbnail_l	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1051-8215&client=summon
thumbnail_m	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1051-8215&client=summon
thumbnail_s	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1051-8215&client=summon