In EDS ansehen

Compression Helps Deep Learning in Image Classification

Gespeichert in:
Bibliographische Detailangaben
Titel: Compression Helps Deep Learning in Image Classification
Autoren: En-Hui Yang, Hossam Amer, Yanbing Jiang
Quelle: Entropy, Vol 23, Iss 881, p 881 (2021)
Verlagsinformationen: MDPI AG
Publikationsjahr: 2021
Bestand: Directory of Open Access Journals: DOAJ Articles
Schlagwörter: image compression, deep learning, inception network, residual network, JPEG, Science, Astrophysics, QB460-466, Physics, QC1-999
Beschreibung: The impact of JPEG compression on deep learning (DL) in image classification is revisited. Given an underlying deep neural network (DNN) pre-trained with pristine ImageNet images, it is demonstrated that, if, for any original image, one can select, among its many JPEG compressed versions including its original version, a suitable version as an input to the underlying DNN, then the classification accuracy of the underlying DNN can be improved significantly while the size in bits of the selected input is, on average, reduced dramatically in comparison with the original image. This is in contrast to the conventional understanding that JPEG compression generally degrades the classification accuracy of DL. Specifically, for each original image, consider its 10 JPEG compressed versions with their quality factor (QF) values from { 100 , 90 , 80 , 70 , 60 , 50 , 40 , 30 , 20 , 10 } . Under the assumption that the ground truth label of the original image is known at the time of selecting an input, but unknown to the underlying DNN, we present a selector called Highest Rank Selector (HRS). It is shown that HRS is optimal in the sense of achieving the highest Top k accuracy on any set of images for any k among all possible selectors. When the underlying DNN is Inception V3 or ResNet-50 V2, HRS improves, on average, the Top 1 classification accuracy and Top 5 classification accuracy on the whole ImageNet validation dataset by 5.6% and 1.9%, respectively, while reducing the input size in bits dramatically—the compression ratio (CR) between the size of the original images and the size of the selected input images by HRS is 8 for the whole ImageNet validation dataset. When the ground truth label of the original image is unknown at the time of selection, we further propose a new convolutional neural network (CNN) topology which is based on the underlying DNN and takes the original image and its 10 JPEG compressed versions as 11 parallel inputs. It is demonstrated that the proposed new CNN ...
Publikationsart: article in journal/newspaper
Sprache: English
Relation: https://www.mdpi.com/1099-4300/23/7/881; https://doaj.org/toc/1099-4300; https://doaj.org/article/7de7ec1ee4b847ff9a21ac06293435ee
DOI: 10.3390/e23070881
Verfügbarkeit: https://doi.org/10.3390/e23070881
https://doaj.org/article/7de7ec1ee4b847ff9a21ac06293435ee
Dokumentencode: edsbas.52F7AA1E
Datenbank: BASE
FullText Text:
  Availability: 0
CustomLinks:
  – Url: https://doi.org/10.3390/e23070881#
    Name: EDS - BASE (s4221598)
    Category: fullText
    Text: View record from BASE
  – Url: https://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=EBSCO&SrcAuth=EBSCO&DestApp=WOS&ServiceName=TransferToWoS&DestLinkType=GeneralSearchSummary&Func=Links&author=Yang%20E
    Name: ISI
    Category: fullText
    Text: Nájsť tento článok vo Web of Science
    Icon: https://imagesrvr.epnet.com/ls/20docs.gif
    MouseOverText: Nájsť tento článok vo Web of Science
Header DbId: edsbas
DbLabel: BASE
An: edsbas.52F7AA1E
RelevancyScore: 919
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 918.9462890625
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Compression Helps Deep Learning in Image Classification
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22En-Hui+Yang%22">En-Hui Yang</searchLink><br /><searchLink fieldCode="AR" term="%22Hossam+Amer%22">Hossam Amer</searchLink><br /><searchLink fieldCode="AR" term="%22Yanbing+Jiang%22">Yanbing Jiang</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: Entropy, Vol 23, Iss 881, p 881 (2021)
– Name: Publisher
  Label: Publisher Information
  Group: PubInfo
  Data: MDPI AG
– Name: DatePubCY
  Label: Publication Year
  Group: Date
  Data: 2021
– Name: Subset
  Label: Collection
  Group: HoldingsInfo
  Data: Directory of Open Access Journals: DOAJ Articles
– Name: Subject
  Label: Subject Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22image+compression%22">image compression</searchLink><br /><searchLink fieldCode="DE" term="%22deep+learning%22">deep learning</searchLink><br /><searchLink fieldCode="DE" term="%22inception+network%22">inception network</searchLink><br /><searchLink fieldCode="DE" term="%22residual+network%22">residual network</searchLink><br /><searchLink fieldCode="DE" term="%22JPEG%22">JPEG</searchLink><br /><searchLink fieldCode="DE" term="%22Science%22">Science</searchLink><br /><searchLink fieldCode="DE" term="%22Astrophysics%22">Astrophysics</searchLink><br /><searchLink fieldCode="DE" term="%22QB460-466%22">QB460-466</searchLink><br /><searchLink fieldCode="DE" term="%22Physics%22">Physics</searchLink><br /><searchLink fieldCode="DE" term="%22QC1-999%22">QC1-999</searchLink>
– Name: Abstract
  Label: Description
  Group: Ab
  Data: The impact of JPEG compression on deep learning (DL) in image classification is revisited. Given an underlying deep neural network (DNN) pre-trained with pristine ImageNet images, it is demonstrated that, if, for any original image, one can select, among its many JPEG compressed versions including its original version, a suitable version as an input to the underlying DNN, then the classification accuracy of the underlying DNN can be improved significantly while the size in bits of the selected input is, on average, reduced dramatically in comparison with the original image. This is in contrast to the conventional understanding that JPEG compression generally degrades the classification accuracy of DL. Specifically, for each original image, consider its 10 JPEG compressed versions with their quality factor (QF) values from <semantics> { 100 , 90 , 80 , 70 , 60 , 50 , 40 , 30 , 20 , 10 } </semantics> . Under the assumption that the ground truth label of the original image is known at the time of selecting an input, but unknown to the underlying DNN, we present a selector called Highest Rank Selector (HRS). It is shown that HRS is optimal in the sense of achieving the highest Top k accuracy on any set of images for any k among all possible selectors. When the underlying DNN is Inception V3 or ResNet-50 V2, HRS improves, on average, the Top 1 classification accuracy and Top 5 classification accuracy on the whole ImageNet validation dataset by 5.6% and 1.9%, respectively, while reducing the input size in bits dramatically—the compression ratio (CR) between the size of the original images and the size of the selected input images by HRS is 8 for the whole ImageNet validation dataset. When the ground truth label of the original image is unknown at the time of selection, we further propose a new convolutional neural network (CNN) topology which is based on the underlying DNN and takes the original image and its 10 JPEG compressed versions as 11 parallel inputs. It is demonstrated that the proposed new CNN ...
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: article in journal/newspaper
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: NoteTitleSource
  Label: Relation
  Group: SrcInfo
  Data: https://www.mdpi.com/1099-4300/23/7/881; https://doaj.org/toc/1099-4300; https://doaj.org/article/7de7ec1ee4b847ff9a21ac06293435ee
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.3390/e23070881
– Name: URL
  Label: Availability
  Group: URL
  Data: https://doi.org/10.3390/e23070881<br />https://doaj.org/article/7de7ec1ee4b847ff9a21ac06293435ee
– Name: AN
  Label: Accession Number
  Group: ID
  Data: edsbas.52F7AA1E
PLink https://erproxy.cvtisr.sk/sfx/access?url=https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=edsbas&AN=edsbas.52F7AA1E
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.3390/e23070881
    Languages:
      – Text: English
    Subjects:
      – SubjectFull: image compression
        Type: general
      – SubjectFull: deep learning
        Type: general
      – SubjectFull: inception network
        Type: general
      – SubjectFull: residual network
        Type: general
      – SubjectFull: JPEG
        Type: general
      – SubjectFull: Science
        Type: general
      – SubjectFull: Astrophysics
        Type: general
      – SubjectFull: QB460-466
        Type: general
      – SubjectFull: Physics
        Type: general
      – SubjectFull: QC1-999
        Type: general
    Titles:
      – TitleFull: Compression Helps Deep Learning in Image Classification
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: En-Hui Yang
      – PersonEntity:
          Name:
            NameFull: Hossam Amer
      – PersonEntity:
          Name:
            NameFull: Yanbing Jiang
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2021
          Identifiers:
            – Type: issn-locals
              Value: edsbas
            – Type: issn-locals
              Value: edsbas.oa
          Titles:
            – TitleFull: Entropy, Vol 23, Iss 881, p 881 (2021
              Type: main
ResultId 1