Advanced Implementation of DNN Translator using ResNet9 for Edge Devices

Resource limitations remain challenging in designing and implementing Deep Neural Networks (DNNs) on edge devices. The high complexity of DNN architectures and the development of these models using hardware languages require high-level verification to ensure they run on specific edge devices such as...

Full description

Saved in:
Bibliographic Details
Published inInternational Journal of Networking and Computing Vol. 14; no. 2; pp. 145 - 156
Main Authors Morishita, Yuki, Amagasaki, Motoki, Kiyama, Masato, Ito, Masayoshi, Diana, Mery
Format Journal Article
LanguageEnglish
Published IJNC Editorial Committee 2024
Subjects
Online AccessGet full text
ISSN2185-2839
2185-2847
2185-2847
DOI10.15803/ijnc.14.2_145

Cover

Abstract Resource limitations remain challenging in designing and implementing Deep Neural Networks (DNNs) on edge devices. The high complexity of DNN architectures and the development of these models using hardware languages require high-level verification to ensure they run on specific edge devices such as FPGA (Field Programmable Gate Array). To address these issues, the DNN translator was developed and performed well in the basic models such as MLP (Multi-layer Perceptron) and LeNet5. The DNN translator generates the DNN models and their parameters for performing the High-Level Synthesis or HLS technology in C++. In this study, we applied ResNet as a DNN model with more complex architecture from the CNNs (Convolutional Neural Networks) family. As a result, the generated C++ files for the ResNet9 and its weights successfully underwent synthesis and implementation on FPGA (Arty A7-100) using Vitis HLS.
AbstractList Resource limitations remain challenging in designing and implementing Deep Neural Networks (DNNs) on edge devices. The high complexity of DNN architectures and the development of these models using hardware languages require high-level verification to ensure they run on specific edge devices such as FPGA (Field Programmable Gate Array). To address these issues, the DNN translator was developed and performed well in the basic models such as MLP (Multi-layer Perceptron) and LeNet5. The DNN translator generates the DNN models and their parameters for performing the High-Level Synthesis or HLS technology in C++. In this study, we applied ResNet as a DNN model with more complex architecture from the CNNs (Convolutional Neural Networks) family. As a result, the generated C++ files for the ResNet9 and its weights successfully underwent synthesis and implementation on FPGA (Arty A7-100) using Vitis HLS.
Author Ito, Masayoshi
Morishita, Yuki
Diana, Mery
Kiyama, Masato
Amagasaki, Motoki
Author_xml – sequence: 1
  fullname: Morishita, Yuki
  organization: Mitsubishi Electric Engineering
– sequence: 1
  fullname: Amagasaki, Motoki
  organization: Kumamoto University
– sequence: 1
  fullname: Kiyama, Masato
  organization: Kumamoto University
– sequence: 1
  fullname: Ito, Masayoshi
  organization: Mitsubishi Electric Engineering
– sequence: 1
  fullname: Diana, Mery
  organization: Kumamoto University
BookMark eNqFkE1rAjEQhkOxUGu99pw_sOtukv3IpSBqqyAWij0vs8nErqxZSVaL_76rFqGnzmWGmfeZw_NIeraxSMhzHIVxkkd8VG2tCmMRsiIWyR3pszhPApaLrHebuXwgQ--3UVdZJiPG-2Q-1kewCjVd7PY17tC20FaNpY2h09WKrh1YX0PbOHrwld3QD_QrbCU13WamN0ineKwU-idyb6D2OPztA_L5OltP5sHy_W0xGS8DxaRMAohMWuac6VJo0KWBjAkEU2apNGUpIE9KgVKb2DBI8kwYo1HlOuWp1AKk5AMyuv492D2cvqGui72rduBORRwVFxfF2UWnobi46IjwSijXeO_Q_A-8XIGtb2GDtzi4tlI1_kmfgdtBfYEr0PIf7_V7jg
Cites_doi 10.1145/3495531
10.1109/MSP.2012.2211477
10.1016/j.procs.2020.07.006
10.1145/3204919.3204929
10.3390/electronics9122200
10.1109/DSN-W50199.2020.00014
10.1109/ICCAD.2017.8203809
10.1109/TCAD.2017.2785257
10.1145/3358182
10.1109/TCAD.2015.2513673
10.1145/2789116.2789145
10.1109/CANDARW60564.2023.00017
10.1145/2539036.2539048
10.1145/3486674
ContentType Journal Article
Copyright 2024 International Journal of Networking and Computing
Copyright_xml – notice: 2024 International Journal of Networking and Computing
DBID AAYXX
CITATION
ADTOC
UNPAY
DOI 10.15803/ijnc.14.2_145
DatabaseName CrossRef
Unpaywall for CDI: Periodical Content
Unpaywall
DatabaseTitle CrossRef
DatabaseTitleList
Database_xml – sequence: 1
  dbid: UNPAY
  name: Unpaywall
  url: https://proxy.k.utb.cz/login?url=https://unpaywall.org/
  sourceTypes: Open Access Repository
DeliveryMethod fulltext_linktorsrc
Discipline Computer Science
EISSN 2185-2847
EndPage 156
ExternalDocumentID 10.15803/ijnc.14.2_145
10_15803_ijnc_14_2_145
article_ijnc_14_2_14_145_article_char_en
GroupedDBID 7.U
ALMA_UNASSIGNED_HOLDINGS
JSF
JSH
KQ8
KWQ
OK1
RJT
RZJ
AAYXX
CITATION
ISHAI
ADTOC
UNPAY
ID FETCH-LOGICAL-c2995-a0f6b832db4dadbfa724eafb769fbb4a85b4e9df1f2a5874ffdec8d6369d4a993
IEDL.DBID UNPAY
ISSN 2185-2839
2185-2847
IngestDate Sun Sep 07 11:18:17 EDT 2025
Wed Oct 01 01:51:15 EDT 2025
Wed Sep 03 06:30:39 EDT 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 2
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c2995-a0f6b832db4dadbfa724eafb769fbb4a85b4e9df1f2a5874ffdec8d6369d4a993
OpenAccessLink https://proxy.k.utb.cz/login?url=https://doi.org/10.15803/ijnc.14.2_145
PageCount 12
ParticipantIDs unpaywall_primary_10_15803_ijnc_14_2_145
crossref_primary_10_15803_ijnc_14_2_145
jstage_primary_article_ijnc_14_2_14_145_article_char_en
ProviderPackageCode CITATION
AAYXX
PublicationCentury 2000
PublicationDate 2024
2024-00-00
PublicationDateYYYYMMDD 2024-01-01
PublicationDate_xml – year: 2024
  text: 2024
PublicationDecade 2020
PublicationTitle International Journal of Networking and Computing
PublicationTitleAlternate IJNC
PublicationYear 2024
Publisher IJNC Editorial Committee
Publisher_xml – name: IJNC Editorial Committee
References [12] Javier Mendez, Kay Bierzynski, M. P. Cuéllar, and Diego P. Morales. Edge intelligence: Concepts, architectures, applications, and future directions. ACM Trans. Embed. Comput. Syst., 21(5), oct 2022.
[18] Dianlei Xu, Tong Li, Yong Li, Xiang Su, Sasu Tarkoma, Tao Jiang, Jon Crowcroft, and Pan Hui. Edge Intelligence: Architectures, Challenges, and Applications, June 2020. arXiv:2003.12172 [cs].
[10] Marcos T. Leipnitz and Gabriel L. Nazar. High-level synthesis of approximate designs under real-time constraints. ACM Trans. Embed. Comput. Syst., 18(5s), oct 2019.
[13] Razvan Nane, Vlad-Mihai Sima, Christian Pilato, Jongsok Choi, Blair Fort, Andrew Canis, Yu Ting Chen, Hsuan Hsiao, Stephen Brown, Fabrizio Ferrandi, Jason Anderson, and Koen Bertels. A survey and evaluation of fpga high-level synthesis tools. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 35(10):1591–1604, 2016.
[17] Xilinx, Inc. Xilinx vitis high-level synthesis (hls) documentation, 2022. accessed May 14.
[1] Donald G. Bailey. The advantages and limitations of high level synthesis for fpga based image processing. In Proceedings of the 9th International Conference on Distributed Smart Cameras, ICDSC '15, page 134–139, 2015.
[6] Alireza Ghaffari and Yvon Savaria. Cnn2gate: An implementation of convolutional neural networks inference on fpgas with automated design space exploration. Electronics, 9(12), 2020.
[9] A. Krizhevsky and G. Hinton. Learning multiple layers of features from tiny images. Master's thesis, Department of Computer Science, University of Toronto, 2009.
[11] A. Mahmoud, N. Aggarwal, A. Nobbe, J. R. S. Vicarte, S. V. Adve, C. W. Fletcher, I. Frosio, and S. K. S. Hari. Pytorchfi: A runtime perturbation tool for dnns. In Proceeding of 2020 50th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshops (DSN-W), pages 25–31, 2020.
[8] Andreas Kloeckner and Contributors. cgen -code generation library. https://github.com/inducer/cgen, 2020. accessed May 12.
[4] Li Deng. The mnist database of handwritten digit images for machine learning research. IEEE Signal Processing Magazine, 29(6):141–142, 2012.
[16] Zi Wang and Benjamin Carrion Schafer. Learning from the past: Efficient high-level synthesis design space exploration for fpgas. ACM Trans. Des. Autom. Electron. Syst., 27(4), 2022.
[2] Yosi Ben-Asher and Nadav Rotem. The benefits of using variable-length pipelined operations in high-level synthesis. ACM Trans. Embed. Comput. Syst., 13(3), dec 2013.
[20] Jieru Zhao, Liang Feng, Sharad Sinha, Wei Zhang, Yun Liang, and Bingsheng He. Comba: A comprehensive model-based analysis framework for high level synthesis of real applications. In Proceeding of 2017 IEEE/ACM International Conference on Computer-Aided Design (ICCAD), pages 430–437, 2017.
[3] Olivier Debauche, Saïd Mahmoudi, Sidi Ahmed Mahmoudi, Pierre Manneback, and Frédéric Lebeau. A new edge architecture for ai-iot services deployment. Procedia Computer Science, 175:10–19, 2020.
[15] Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, and Soumith Chintala. Pytorch: An imperative style, high-performance deep learning library. https://pytorch.org, 2021.
[7] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition, 2015.
[5] Diana, Mery, Kiyama, Masato, Amagasaki, Motoki, Ito, Masayoshi, and Morishita, Yuki. Deep neural network translator for edge site implementation. In Proceeding of 2023 Eleventh International Symposium on Computing and Networking Workshops (CANDARW), page 52, Matsue, Japan, nov 2023.
[14] Katharina Ostaszewski, Philip Heinisch, and Hendrik Ranocha. Advantages and pitfalls of opencl in computational physics. In Proceedings of the International Workshop on OpenCL, IWOCL '18, New York, NY, USA, 2018.
[19] Chen Zhang, Guangyu Sun, Zhenman Fang, Peipei Zhou, Peichen Pan, and Jason Cong. Caffeine: Toward uniformed representation and acceleration for deep convolutional neural networks. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 38(11):2072–2085, 2019.
11
12
13
14
15
16
17
18
19
1
2
3
4
5
6
7
8
9
20
10
References_xml – reference: [4] Li Deng. The mnist database of handwritten digit images for machine learning research. IEEE Signal Processing Magazine, 29(6):141–142, 2012.
– reference: [11] A. Mahmoud, N. Aggarwal, A. Nobbe, J. R. S. Vicarte, S. V. Adve, C. W. Fletcher, I. Frosio, and S. K. S. Hari. Pytorchfi: A runtime perturbation tool for dnns. In Proceeding of 2020 50th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshops (DSN-W), pages 25–31, 2020.
– reference: [15] Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, and Soumith Chintala. Pytorch: An imperative style, high-performance deep learning library. https://pytorch.org, 2021.
– reference: [20] Jieru Zhao, Liang Feng, Sharad Sinha, Wei Zhang, Yun Liang, and Bingsheng He. Comba: A comprehensive model-based analysis framework for high level synthesis of real applications. In Proceeding of 2017 IEEE/ACM International Conference on Computer-Aided Design (ICCAD), pages 430–437, 2017.
– reference: [14] Katharina Ostaszewski, Philip Heinisch, and Hendrik Ranocha. Advantages and pitfalls of opencl in computational physics. In Proceedings of the International Workshop on OpenCL, IWOCL '18, New York, NY, USA, 2018.
– reference: [3] Olivier Debauche, Saïd Mahmoudi, Sidi Ahmed Mahmoudi, Pierre Manneback, and Frédéric Lebeau. A new edge architecture for ai-iot services deployment. Procedia Computer Science, 175:10–19, 2020.
– reference: [16] Zi Wang and Benjamin Carrion Schafer. Learning from the past: Efficient high-level synthesis design space exploration for fpgas. ACM Trans. Des. Autom. Electron. Syst., 27(4), 2022.
– reference: [19] Chen Zhang, Guangyu Sun, Zhenman Fang, Peipei Zhou, Peichen Pan, and Jason Cong. Caffeine: Toward uniformed representation and acceleration for deep convolutional neural networks. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 38(11):2072–2085, 2019.
– reference: [1] Donald G. Bailey. The advantages and limitations of high level synthesis for fpga based image processing. In Proceedings of the 9th International Conference on Distributed Smart Cameras, ICDSC '15, page 134–139, 2015.
– reference: [17] Xilinx, Inc. Xilinx vitis high-level synthesis (hls) documentation, 2022. accessed May 14.
– reference: [18] Dianlei Xu, Tong Li, Yong Li, Xiang Su, Sasu Tarkoma, Tao Jiang, Jon Crowcroft, and Pan Hui. Edge Intelligence: Architectures, Challenges, and Applications, June 2020. arXiv:2003.12172 [cs].
– reference: [8] Andreas Kloeckner and Contributors. cgen -code generation library. https://github.com/inducer/cgen, 2020. accessed May 12.
– reference: [13] Razvan Nane, Vlad-Mihai Sima, Christian Pilato, Jongsok Choi, Blair Fort, Andrew Canis, Yu Ting Chen, Hsuan Hsiao, Stephen Brown, Fabrizio Ferrandi, Jason Anderson, and Koen Bertels. A survey and evaluation of fpga high-level synthesis tools. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 35(10):1591–1604, 2016.
– reference: [5] Diana, Mery, Kiyama, Masato, Amagasaki, Motoki, Ito, Masayoshi, and Morishita, Yuki. Deep neural network translator for edge site implementation. In Proceeding of 2023 Eleventh International Symposium on Computing and Networking Workshops (CANDARW), page 52, Matsue, Japan, nov 2023.
– reference: [2] Yosi Ben-Asher and Nadav Rotem. The benefits of using variable-length pipelined operations in high-level synthesis. ACM Trans. Embed. Comput. Syst., 13(3), dec 2013.
– reference: [9] A. Krizhevsky and G. Hinton. Learning multiple layers of features from tiny images. Master's thesis, Department of Computer Science, University of Toronto, 2009.
– reference: [10] Marcos T. Leipnitz and Gabriel L. Nazar. High-level synthesis of approximate designs under real-time constraints. ACM Trans. Embed. Comput. Syst., 18(5s), oct 2019.
– reference: [7] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition, 2015.
– reference: [12] Javier Mendez, Kay Bierzynski, M. P. Cuéllar, and Diego P. Morales. Edge intelligence: Concepts, architectures, applications, and future directions. ACM Trans. Embed. Comput. Syst., 21(5), oct 2022.
– reference: [6] Alireza Ghaffari and Yvon Savaria. Cnn2gate: An implementation of convolutional neural networks inference on fpgas with automated design space exploration. Electronics, 9(12), 2020.
– ident: 17
– ident: 18
– ident: 16
  doi: 10.1145/3495531
– ident: 4
  doi: 10.1109/MSP.2012.2211477
– ident: 3
  doi: 10.1016/j.procs.2020.07.006
– ident: 14
  doi: 10.1145/3204919.3204929
– ident: 6
  doi: 10.3390/electronics9122200
– ident: 11
  doi: 10.1109/DSN-W50199.2020.00014
– ident: 20
  doi: 10.1109/ICCAD.2017.8203809
– ident: 15
– ident: 19
  doi: 10.1109/TCAD.2017.2785257
– ident: 10
  doi: 10.1145/3358182
– ident: 13
  doi: 10.1109/TCAD.2015.2513673
– ident: 1
  doi: 10.1145/2789116.2789145
– ident: 5
  doi: 10.1109/CANDARW60564.2023.00017
– ident: 9
– ident: 2
  doi: 10.1145/2539036.2539048
– ident: 7
– ident: 8
– ident: 12
  doi: 10.1145/3486674
SSID ssj0000779023
Score 2.2451918
Snippet Resource limitations remain challenging in designing and implementing Deep Neural Networks (DNNs) on edge devices. The high complexity of DNN architectures and...
SourceID unpaywall
crossref
jstage
SourceType Open Access Repository
Index Database
Publisher
StartPage 145
SubjectTerms Deep Neural Network Translator
Edge Site
High-Level Synthesis
ResNet9
Title Advanced Implementation of DNN Translator using ResNet9 for Edge Devices
URI https://www.jstage.jst.go.jp/article/ijnc/14/2/14_145/_article/-char/en
https://doi.org/10.15803/ijnc.14.2_145
UnpaywallVersion publishedVersion
Volume 14
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
ispartofPNX International Journal of Networking and Computing, 2024, Vol.14(2), pp.145-156
journalDatabaseRights – providerCode: PRVAFT
  databaseName: Open Access Digital Library
  customDbUrl:
  eissn: 2185-2847
  dateEnd: 99991231
  omitProxy: true
  ssIdentifier: ssj0000779023
  issn: 2185-2847
  databaseCode: KQ8
  dateStart: 20110101
  isFulltext: true
  titleUrlDefault: http://grweb.coalliance.org/oadl/oadl.html
  providerName: Colorado Alliance of Research Libraries
link http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwjV1LSwMxEB5sK3iyPrGikoOgl63tbjabPRZfRXFRsVBPJY9JUetatEX015ts0_o66G1hh90wyeT7QuabAdhNtDIy4iqgcYIB1c1m4HAnsMitQi4RUTjt8EXG2h161o27PkHWaWG-3t_HvBEd3N3bOGnSemhJalyCCost5y5DpZNdtm5d5zgLOIHFyPTzmSa-OuPvD3xDn_l7S8Ccsn5hnA_F26sYDL5gy0kVTqejmqSUPNTHI1lX7z8KNv497CVY9PSStCbrYRnmMF-B6rR1A_GRvArtlr_7J0V54EevQMrJkyFHWUYKCBu4AzlxmfF9co0vGY5SYjkuOdZ9JEdY7DFr0Dk5vjlsB76pQqBCp8YWDcOkDWMtqRZaGpGEFIWRCUuNlFTwWFJMtWmaUMQ8ocZoVFyziKWaCstm1qGcP-W4AQS5ibVinDNltwIdScbRGCZCTCzpVKoGe1Nn94aT2hk9d-ZwDuo5B1nf9AoH1SCZzMXMzsfNNzNnOXvhRGk2smuwP5u8P36y-X_TLSiPnse4bfnGSO5A6fyK7_gl9wEvQ9Y5
linkProvider Unpaywall
linkToUnpaywall http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwjV1NTwIxEG0UTDyJnxGjpgcTvSyyu223eyQCEhM3xkiCJ9KPqRFxIQox-uttoSDoAW-b7GS3mXb6XtN5MwidJVoZGXMVEJpAQHQYBg53AovcKuISAITTDt9mrNUmNx3a8QmyTguzeH9PeTW-fO7ZOAlJJbIkla6jIqOWcxdQsZ3d1R5d5zgLOIHFyPTnmSS-OuPfDyyhz0bPEjCnrN8c50Px-SH6_QVsaZbQ9WxU05SSl8p4JCvq61fBxtXD3kZbnl7i2nQ97KA1yHdRada6AftI3kOtmr_7x5PywK9egZTjgcH1LMMTCOu7Azl2mfFP-B7eMxil2HJc3NBPgOsw2WP2UbvZeLhqBb6pQqAip8YWVcOkDWMtiRZaGpFEBISRCUuNlERwKgmk2oQmEpQnxBgNimsWs1QTYdnMASrkgxwOEQZuqFaMc6bsVqBjyTgYw0QEiSWdSpXR-czZ3eG0dkbXnTmcg7rOQdY33YmDyiiZzsXczsfNkpmznL9wojQb2WV0MZ-8FT85-r_pMSqM3sZwYvnGSJ76xfYNL13VRA
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=Advanced+Implementation+of+DNN+Translator+using+ResNet9+for+Edge+Devices&rft.jtitle=International+Journal+of+Networking+and+Computing&rft.au=Morishita%2C+Yuki&rft.au=Amagasaki%2C+Motoki&rft.au=Kiyama%2C+Masato&rft.au=Ito%2C+Masayoshi&rft.date=2024&rft.pub=IJNC+Editorial+Committee&rft.issn=2185-2839&rft.eissn=2185-2847&rft.volume=14&rft.issue=2&rft.spage=145&rft.epage=156&rft_id=info:doi/10.15803%2Fijnc.14.2_145&rft.externalDocID=article_ijnc_14_2_14_145_article_char_en
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2185-2839&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2185-2839&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2185-2839&client=summon