Robust convolutional neural networks in sram-based fpgas: A case study in image classification

Página do item simplificado
dc.contributor.authorBENEVENUTI, F.
dc.contributor.authorKASTENSMIDT, F.
dc.contributor.authorOLIVEIRA, A.
dc.contributor.authorADDED, N.
dc.contributor.authorAGUIAR, V.
dc.contributor.authorMEDINA, N.
dc.contributor.authorMarcilei Aparecida Guazzelli
dc.contributor.authorOrcidhttps://orcid.org/0000-0001-7110-7241
dc.date.accessioned2022-01-12T21:53:46Z
dc.date.available2022-01-12T21:53:46Z
dc.date.issued2021-08-23
dc.description.abstract© 2021, Brazilian Microelectronics Society. All rights reserved.— This work discusses the main aspects of vulnerability and degradation of accuracy of an image classification engine implemented into SRAM-based FPGAs under faults. The image classification engine is an all-convolutional neural-network (CNN) trained with a dataset of traffic sign recognition benchmark. The Caffe and Ristretto frameworks were used for CNN training and fine-tuning while the ZynqNet inference engine was adopted as hardware implementation on a Xilinx 28 nm SRAM-based FPGA. The CNN under test was generated using an evolutive approach based on genetic algorithm. The methodologies for qualifying this CNN under faults is presented and both heavy-ions accelerated irradiation and emulated fault injection were performed. To cross validate results from radiation and fault injection, different implementations of the same CNN were tested using reduced arithmetic precision and protection of user data by Hamming codes, in combination with configuration memory healing by the scrubbing mechanism available in Xilinx FPGA. Some of these alternative implementations increased significantly the mission time of the CNN, when compared to the original ZynqNet operating on 32 bits floating point number, and the experiment suggests areas for further improvements on the fault injection methodology in use.
dc.description.issuenumber2
dc.description.volume16
dc.identifier.citationBENEVENUTI, F.; KASTENSMIDT, F.; OLIVEIRA, A.; ADDED, N.; AGUIAR, V.; MEDINA, N.; GUAZZELLI, M. A. Robust convolutional neural networks in sram-based fpgas: A case study in image classification. Journal of Integrated Circuits and Systems, v. 16, n. 2, Aug. 2021.
dc.identifier.doi10.29292/jics.v16i2.504
dc.identifier.issn1872-0234
dc.identifier.urihttps://repositorio.fei.edu.br/handle/FEI/3562
dc.relation.ispartofJournal of Integrated Circuits and Systems
dc.rightsAcesso Aberto
dc.subject.otherlanguageDeep Learning
dc.subject.otherlanguageNeural Networks
dc.subject.otherlanguageReliability
dc.subject.otherlanguageSingle-event Effects
dc.subject.otherlanguageSRAM-based FPGA
dc.titleRobust convolutional neural networks in sram-based fpgas: A case study in image classification
dc.typeArtigo
dcterms.licenseOpen Journal Systems "Este é um artigo publicado em acesso aberto sob uma licença de código aberto (GPL v2). Fonte: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85114037402&origin=inward. Acesso em: 09 setembro, 2022.
fei.scopus.citations3
fei.scopus.eid2-s2.0-85114037402
fei.scopus.updated2024-11-01
fei.scopus.urlhttps://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85114037402&origin=inward
Arquivos
Pacote Original
Agora exibindo 1 - 1 de 1
Nenhuma Miniatura disponível
Nome:
Guazzelli_pdf
Tamanho:
1.1 MB
Formato:
Adobe Portable Document Format
Baixar
Coleções
Artigos