Studying the plasticity in deep convolutional neural networks using random pruning View Full Text


Ontology type: schema:ScholarlyArticle      Open Access: True


Article Info

DATE

2019-03

AUTHORS

Deepak Mittal, Shweta Bhardwaj, Mitesh M. Khapra, Balaraman Ravindran

ABSTRACT

Recently, there has been a lot of work on pruning filters from deep convolutional neural networks (CNNs) with the intention of reducing computations. The key idea is to rank the filters based on a certain criterion (say, l1-norm, average percentage of zeros, etc.) and retain only the top-ranked filters. Once the low-scoring filters are pruned away, the remainder of the network is fine-tuned and is shown to give performance comparable to the original unpruned network. In this work, we report experiments which suggest that the comparable performance of the pruned network is not due to the specific criterion chosen, but due to the inherent plasticity of deep neural networks which allows them to recover from the loss of pruned filters once the rest of the filters are fine-tuned. Specifically, we show counterintuitive results wherein by randomly pruning 25–50% filters from deep CNNs we are able to obtain the same performance as obtained by using state-of-the-art pruning methods. We empirically validate our claims by doing an exhaustive evaluation with VGG-16 and ResNet-50. Further, we also evaluate a real-world scenario where a CNN trained on all 1000 ImageNet classes needs to be tested on only a small set of classes at test time (say, only animals). We create a new benchmark dataset from ImageNet to evaluate such class-specific pruning and show that even here a random pruning strategy gives close to state-of-the-art performance. Lastly, unlike existing approaches which mainly focus on the task of image classification, in this work we also report results on object detection and image segmentation. We show that using a simple random pruning strategy, we can achieve significant speedup in object detection (74% improvement in fps) while retaining the same accuracy as that of the original Faster-RCNN model. Similarly, we show that the performance of a pruned segmentation network is actually very similar to that of the original unpruned SegNet. More... »

PAGES

203-216

References to SciGraph publications

  • 2016. XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks in COMPUTER VISION – ECCV 2016
  • 2015-12. ImageNet Large Scale Visual Recognition Challenge in INTERNATIONAL JOURNAL OF COMPUTER VISION
  • 2010-06. The Pascal Visual Object Classes (VOC) Challenge in INTERNATIONAL JOURNAL OF COMPUTER VISION
  • 2007. Optimal Brain Surgeon for General Dynamic Neural Networks in PROGRESS IN ARTIFICIAL INTELLIGENCE
  • 2016. SSD: Single Shot MultiBox Detector in COMPUTER VISION – ECCV 2016
  • Identifiers

    URI

    http://scigraph.springernature.com/pub.10.1007/s00138-018-01001-9

    DOI

    http://dx.doi.org/10.1007/s00138-018-01001-9

    DIMENSIONS

    https://app.dimensions.ai/details/publication/pub.1111570790


    Indexing Status Check whether this publication has been indexed by Scopus and Web Of Science using the SN Indexing Status Tool
    Incoming Citations Browse incoming citations for this publication using opencitations.net

    JSON-LD is the canonical representation for SciGraph data.

    TIP: You can open this SciGraph record using an external JSON-LD service: JSON-LD Playground Google SDTT

    [
      {
        "@context": "https://springernature.github.io/scigraph/jsonld/sgcontext.json", 
        "about": [
          {
            "id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/0801", 
            "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
            "name": "Artificial Intelligence and Image Processing", 
            "type": "DefinedTerm"
          }, 
          {
            "id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/08", 
            "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
            "name": "Information and Computing Sciences", 
            "type": "DefinedTerm"
          }
        ], 
        "author": [
          {
            "affiliation": {
              "alternateName": "Indian Institute of Technology Madras", 
              "id": "https://www.grid.ac/institutes/grid.417969.4", 
              "name": [
                "Department of Computer Science and Engineering, Robert Bosch Centre for Data Science and AI (RBC-DSAI), Indian Institute of Technology Madras, Chennai, India"
              ], 
              "type": "Organization"
            }, 
            "familyName": "Mittal", 
            "givenName": "Deepak", 
            "id": "sg:person.012242306142.47", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012242306142.47"
            ], 
            "type": "Person"
          }, 
          {
            "affiliation": {
              "alternateName": "Indian Institute of Technology Madras", 
              "id": "https://www.grid.ac/institutes/grid.417969.4", 
              "name": [
                "Department of Computer Science and Engineering, Robert Bosch Centre for Data Science and AI (RBC-DSAI), Indian Institute of Technology Madras, Chennai, India"
              ], 
              "type": "Organization"
            }, 
            "familyName": "Bhardwaj", 
            "givenName": "Shweta", 
            "id": "sg:person.014454525376.34", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.014454525376.34"
            ], 
            "type": "Person"
          }, 
          {
            "affiliation": {
              "alternateName": "Indian Institute of Technology Madras", 
              "id": "https://www.grid.ac/institutes/grid.417969.4", 
              "name": [
                "Department of Computer Science and Engineering, Robert Bosch Centre for Data Science and AI (RBC-DSAI), Indian Institute of Technology Madras, Chennai, India"
              ], 
              "type": "Organization"
            }, 
            "familyName": "Khapra", 
            "givenName": "Mitesh M.", 
            "id": "sg:person.01233542565.06", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01233542565.06"
            ], 
            "type": "Person"
          }, 
          {
            "affiliation": {
              "alternateName": "Indian Institute of Technology Madras", 
              "id": "https://www.grid.ac/institutes/grid.417969.4", 
              "name": [
                "Department of Computer Science and Engineering, Robert Bosch Centre for Data Science and AI (RBC-DSAI), Indian Institute of Technology Madras, Chennai, India"
              ], 
              "type": "Organization"
            }, 
            "familyName": "Ravindran", 
            "givenName": "Balaraman", 
            "id": "sg:person.015501102741.39", 
            "sameAs": [
              "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.015501102741.39"
            ], 
            "type": "Person"
          }
        ], 
        "citation": [
          {
            "id": "https://doi.org/10.1016/j.patrec.2008.04.005", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1001414384"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/978-3-540-77002-2_2", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1004228400", 
              "https://doi.org/10.1007/978-3-540-77002-2_2"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/978-3-540-77002-2_2", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1004228400", 
              "https://doi.org/10.1007/978-3-540-77002-2_2"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/978-3-319-46493-0_32", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1005632531", 
              "https://doi.org/10.1007/978-3-319-46493-0_32"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/s11263-015-0816-y", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1009767488", 
              "https://doi.org/10.1007/s11263-015-0816-y"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/s11263-009-0275-4", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1014796149", 
              "https://doi.org/10.1007/s11263-009-0275-4"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/s11263-009-0275-4", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1014796149", 
              "https://doi.org/10.1007/s11263-009-0275-4"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "sg:pub.10.1007/978-3-319-46448-0_2", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1017177111", 
              "https://doi.org/10.1007/978-3-319-46448-0_2"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1177/0269215511431903", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1032913631"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1177/0269215511431903", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1032913631"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/tpami.2016.2577031", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1061745117"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/tpami.2016.2644615", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1061745216"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1145/3005348", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1083756204"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/cvpr.2016.90", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1093359587"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/cvpr.2015.7298965", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1093626237"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/cvpr.2015.7298809", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1094001018"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/cvpr.2014.81", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1094727707"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/iccv.2015.169", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1095573598"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/cvpr.2016.91", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1095811486"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/cvpr.2017.195", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1095837190"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/cvpr.2017.243", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1095843442"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.5244/c.28.88", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1099426768"
            ], 
            "type": "CreativeWork"
          }, 
          {
            "id": "https://doi.org/10.1109/iccv.2017.155", 
            "sameAs": [
              "https://app.dimensions.ai/details/publication/pub.1100060121"
            ], 
            "type": "CreativeWork"
          }
        ], 
        "datePublished": "2019-03", 
        "datePublishedReg": "2019-03-01", 
        "description": "Recently, there has been a lot of work on pruning filters from deep convolutional neural networks (CNNs) with the intention of reducing computations. The key idea is to rank the filters based on a certain criterion (say, l1-norm, average percentage of zeros, etc.) and retain only the top-ranked filters. Once the low-scoring filters are pruned away, the remainder of the network is fine-tuned and is shown to give performance comparable to the original unpruned network. In this work, we report experiments which suggest that the comparable performance of the pruned network is not due to the specific criterion chosen, but due to the inherent plasticity of deep neural networks which allows them to recover from the loss of pruned filters once the rest of the filters are fine-tuned. Specifically, we show counterintuitive results wherein by randomly pruning 25\u201350% filters from deep CNNs we are able to obtain the same performance as obtained by using state-of-the-art pruning methods. We empirically validate our claims by doing an exhaustive evaluation with VGG-16 and ResNet-50. Further, we also evaluate a real-world scenario where a CNN trained on all 1000 ImageNet classes needs to be tested on only a small set of classes at test time (say, only animals). We create a new benchmark dataset from ImageNet to evaluate such class-specific pruning and show that even here a random pruning strategy gives close to state-of-the-art performance. Lastly, unlike existing approaches which mainly focus on the task of image classification, in this work we also report results on object detection and image segmentation. We show that using a simple random pruning strategy, we can achieve significant speedup in object detection (74% improvement in fps) while retaining the same accuracy as that of the original Faster-RCNN model. Similarly, we show that the performance of a pruned segmentation network is actually very similar to that of the original unpruned SegNet.", 
        "genre": "research_article", 
        "id": "sg:pub.10.1007/s00138-018-01001-9", 
        "inLanguage": [
          "en"
        ], 
        "isAccessibleForFree": true, 
        "isPartOf": [
          {
            "id": "sg:journal.1045266", 
            "issn": [
              "0932-8092", 
              "1432-1769"
            ], 
            "name": "Machine Vision and Applications", 
            "type": "Periodical"
          }, 
          {
            "issueNumber": "2", 
            "type": "PublicationIssue"
          }, 
          {
            "type": "PublicationVolume", 
            "volumeNumber": "30"
          }
        ], 
        "name": "Studying the plasticity in deep convolutional neural networks using random pruning", 
        "pagination": "203-216", 
        "productId": [
          {
            "name": "readcube_id", 
            "type": "PropertyValue", 
            "value": [
              "9ec107f90abf9248dd30e1d364780ec096e7c2536e6d34e68d1efb5983b8df4b"
            ]
          }, 
          {
            "name": "doi", 
            "type": "PropertyValue", 
            "value": [
              "10.1007/s00138-018-01001-9"
            ]
          }, 
          {
            "name": "dimensions_id", 
            "type": "PropertyValue", 
            "value": [
              "pub.1111570790"
            ]
          }
        ], 
        "sameAs": [
          "https://doi.org/10.1007/s00138-018-01001-9", 
          "https://app.dimensions.ai/details/publication/pub.1111570790"
        ], 
        "sdDataset": "articles", 
        "sdDatePublished": "2019-04-11T12:37", 
        "sdLicense": "https://scigraph.springernature.com/explorer/license/", 
        "sdPublisher": {
          "name": "Springer Nature - SN SciGraph project", 
          "type": "Organization"
        }, 
        "sdSource": "s3://com-uberresearch-data-dimensions-target-20181106-alternative/cleanup/v134/2549eaecd7973599484d7c17b260dba0a4ecb94b/merge/v9/a6c9fde33151104705d4d7ff012ea9563521a3ce/jats-lookup/v90/0000000363_0000000363/records_70032_00000003.jsonl", 
        "type": "ScholarlyArticle", 
        "url": "https://link.springer.com/10.1007%2Fs00138-018-01001-9"
      }
    ]
     

    Download the RDF metadata as:  json-ld nt turtle xml License info

    HOW TO GET THIS DATA PROGRAMMATICALLY:

    JSON-LD is a popular format for linked data which is fully compatible with JSON.

    curl -H 'Accept: application/ld+json' 'https://scigraph.springernature.com/pub.10.1007/s00138-018-01001-9'

    N-Triples is a line-based linked data format ideal for batch operations.

    curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1007/s00138-018-01001-9'

    Turtle is a human-readable linked data format.

    curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/s00138-018-01001-9'

    RDF/XML is a standard XML format for linked data.

    curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/s00138-018-01001-9'


     

    This table displays all metadata directly associated to this object as RDF triples.

    147 TRIPLES      21 PREDICATES      47 URIs      19 LITERALS      7 BLANK NODES

    Subject Predicate Object
    1 sg:pub.10.1007/s00138-018-01001-9 schema:about anzsrc-for:08
    2 anzsrc-for:0801
    3 schema:author N423fd3d8c5484bb59d918e81451a7bc4
    4 schema:citation sg:pub.10.1007/978-3-319-46448-0_2
    5 sg:pub.10.1007/978-3-319-46493-0_32
    6 sg:pub.10.1007/978-3-540-77002-2_2
    7 sg:pub.10.1007/s11263-009-0275-4
    8 sg:pub.10.1007/s11263-015-0816-y
    9 https://doi.org/10.1016/j.patrec.2008.04.005
    10 https://doi.org/10.1109/cvpr.2014.81
    11 https://doi.org/10.1109/cvpr.2015.7298809
    12 https://doi.org/10.1109/cvpr.2015.7298965
    13 https://doi.org/10.1109/cvpr.2016.90
    14 https://doi.org/10.1109/cvpr.2016.91
    15 https://doi.org/10.1109/cvpr.2017.195
    16 https://doi.org/10.1109/cvpr.2017.243
    17 https://doi.org/10.1109/iccv.2015.169
    18 https://doi.org/10.1109/iccv.2017.155
    19 https://doi.org/10.1109/tpami.2016.2577031
    20 https://doi.org/10.1109/tpami.2016.2644615
    21 https://doi.org/10.1145/3005348
    22 https://doi.org/10.1177/0269215511431903
    23 https://doi.org/10.5244/c.28.88
    24 schema:datePublished 2019-03
    25 schema:datePublishedReg 2019-03-01
    26 schema:description Recently, there has been a lot of work on pruning filters from deep convolutional neural networks (CNNs) with the intention of reducing computations. The key idea is to rank the filters based on a certain criterion (say, l1-norm, average percentage of zeros, etc.) and retain only the top-ranked filters. Once the low-scoring filters are pruned away, the remainder of the network is fine-tuned and is shown to give performance comparable to the original unpruned network. In this work, we report experiments which suggest that the comparable performance of the pruned network is not due to the specific criterion chosen, but due to the inherent plasticity of deep neural networks which allows them to recover from the loss of pruned filters once the rest of the filters are fine-tuned. Specifically, we show counterintuitive results wherein by randomly pruning 25–50% filters from deep CNNs we are able to obtain the same performance as obtained by using state-of-the-art pruning methods. We empirically validate our claims by doing an exhaustive evaluation with VGG-16 and ResNet-50. Further, we also evaluate a real-world scenario where a CNN trained on all 1000 ImageNet classes needs to be tested on only a small set of classes at test time (say, only animals). We create a new benchmark dataset from ImageNet to evaluate such class-specific pruning and show that even here a random pruning strategy gives close to state-of-the-art performance. Lastly, unlike existing approaches which mainly focus on the task of image classification, in this work we also report results on object detection and image segmentation. We show that using a simple random pruning strategy, we can achieve significant speedup in object detection (74% improvement in fps) while retaining the same accuracy as that of the original Faster-RCNN model. Similarly, we show that the performance of a pruned segmentation network is actually very similar to that of the original unpruned SegNet.
    27 schema:genre research_article
    28 schema:inLanguage en
    29 schema:isAccessibleForFree true
    30 schema:isPartOf N162986f00e694e4d97154991556d4cc7
    31 Na455a2e716364843ab51ae816f62872a
    32 sg:journal.1045266
    33 schema:name Studying the plasticity in deep convolutional neural networks using random pruning
    34 schema:pagination 203-216
    35 schema:productId N468b5dac8c5545d589995df400629cf4
    36 N8624142aec1740e896824e546483df3a
    37 Ne37752af916945218698ba0b0f8ff47e
    38 schema:sameAs https://app.dimensions.ai/details/publication/pub.1111570790
    39 https://doi.org/10.1007/s00138-018-01001-9
    40 schema:sdDatePublished 2019-04-11T12:37
    41 schema:sdLicense https://scigraph.springernature.com/explorer/license/
    42 schema:sdPublisher N547993fa534c4535bd433c8c0912e252
    43 schema:url https://link.springer.com/10.1007%2Fs00138-018-01001-9
    44 sgo:license sg:explorer/license/
    45 sgo:sdDataset articles
    46 rdf:type schema:ScholarlyArticle
    47 N04bb08ebfab24bc69bfb18eab41adde9 rdf:first sg:person.014454525376.34
    48 rdf:rest Nc5c71a170f7e492e9df5a1d88aea552c
    49 N162986f00e694e4d97154991556d4cc7 schema:volumeNumber 30
    50 rdf:type schema:PublicationVolume
    51 N423fd3d8c5484bb59d918e81451a7bc4 rdf:first sg:person.012242306142.47
    52 rdf:rest N04bb08ebfab24bc69bfb18eab41adde9
    53 N468b5dac8c5545d589995df400629cf4 schema:name dimensions_id
    54 schema:value pub.1111570790
    55 rdf:type schema:PropertyValue
    56 N547993fa534c4535bd433c8c0912e252 schema:name Springer Nature - SN SciGraph project
    57 rdf:type schema:Organization
    58 N8624142aec1740e896824e546483df3a schema:name readcube_id
    59 schema:value 9ec107f90abf9248dd30e1d364780ec096e7c2536e6d34e68d1efb5983b8df4b
    60 rdf:type schema:PropertyValue
    61 Na455a2e716364843ab51ae816f62872a schema:issueNumber 2
    62 rdf:type schema:PublicationIssue
    63 Nc5c71a170f7e492e9df5a1d88aea552c rdf:first sg:person.01233542565.06
    64 rdf:rest Ncac1de65087147a5ab25f5cdd30f0b9c
    65 Ncac1de65087147a5ab25f5cdd30f0b9c rdf:first sg:person.015501102741.39
    66 rdf:rest rdf:nil
    67 Ne37752af916945218698ba0b0f8ff47e schema:name doi
    68 schema:value 10.1007/s00138-018-01001-9
    69 rdf:type schema:PropertyValue
    70 anzsrc-for:08 schema:inDefinedTermSet anzsrc-for:
    71 schema:name Information and Computing Sciences
    72 rdf:type schema:DefinedTerm
    73 anzsrc-for:0801 schema:inDefinedTermSet anzsrc-for:
    74 schema:name Artificial Intelligence and Image Processing
    75 rdf:type schema:DefinedTerm
    76 sg:journal.1045266 schema:issn 0932-8092
    77 1432-1769
    78 schema:name Machine Vision and Applications
    79 rdf:type schema:Periodical
    80 sg:person.012242306142.47 schema:affiliation https://www.grid.ac/institutes/grid.417969.4
    81 schema:familyName Mittal
    82 schema:givenName Deepak
    83 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012242306142.47
    84 rdf:type schema:Person
    85 sg:person.01233542565.06 schema:affiliation https://www.grid.ac/institutes/grid.417969.4
    86 schema:familyName Khapra
    87 schema:givenName Mitesh M.
    88 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01233542565.06
    89 rdf:type schema:Person
    90 sg:person.014454525376.34 schema:affiliation https://www.grid.ac/institutes/grid.417969.4
    91 schema:familyName Bhardwaj
    92 schema:givenName Shweta
    93 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.014454525376.34
    94 rdf:type schema:Person
    95 sg:person.015501102741.39 schema:affiliation https://www.grid.ac/institutes/grid.417969.4
    96 schema:familyName Ravindran
    97 schema:givenName Balaraman
    98 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.015501102741.39
    99 rdf:type schema:Person
    100 sg:pub.10.1007/978-3-319-46448-0_2 schema:sameAs https://app.dimensions.ai/details/publication/pub.1017177111
    101 https://doi.org/10.1007/978-3-319-46448-0_2
    102 rdf:type schema:CreativeWork
    103 sg:pub.10.1007/978-3-319-46493-0_32 schema:sameAs https://app.dimensions.ai/details/publication/pub.1005632531
    104 https://doi.org/10.1007/978-3-319-46493-0_32
    105 rdf:type schema:CreativeWork
    106 sg:pub.10.1007/978-3-540-77002-2_2 schema:sameAs https://app.dimensions.ai/details/publication/pub.1004228400
    107 https://doi.org/10.1007/978-3-540-77002-2_2
    108 rdf:type schema:CreativeWork
    109 sg:pub.10.1007/s11263-009-0275-4 schema:sameAs https://app.dimensions.ai/details/publication/pub.1014796149
    110 https://doi.org/10.1007/s11263-009-0275-4
    111 rdf:type schema:CreativeWork
    112 sg:pub.10.1007/s11263-015-0816-y schema:sameAs https://app.dimensions.ai/details/publication/pub.1009767488
    113 https://doi.org/10.1007/s11263-015-0816-y
    114 rdf:type schema:CreativeWork
    115 https://doi.org/10.1016/j.patrec.2008.04.005 schema:sameAs https://app.dimensions.ai/details/publication/pub.1001414384
    116 rdf:type schema:CreativeWork
    117 https://doi.org/10.1109/cvpr.2014.81 schema:sameAs https://app.dimensions.ai/details/publication/pub.1094727707
    118 rdf:type schema:CreativeWork
    119 https://doi.org/10.1109/cvpr.2015.7298809 schema:sameAs https://app.dimensions.ai/details/publication/pub.1094001018
    120 rdf:type schema:CreativeWork
    121 https://doi.org/10.1109/cvpr.2015.7298965 schema:sameAs https://app.dimensions.ai/details/publication/pub.1093626237
    122 rdf:type schema:CreativeWork
    123 https://doi.org/10.1109/cvpr.2016.90 schema:sameAs https://app.dimensions.ai/details/publication/pub.1093359587
    124 rdf:type schema:CreativeWork
    125 https://doi.org/10.1109/cvpr.2016.91 schema:sameAs https://app.dimensions.ai/details/publication/pub.1095811486
    126 rdf:type schema:CreativeWork
    127 https://doi.org/10.1109/cvpr.2017.195 schema:sameAs https://app.dimensions.ai/details/publication/pub.1095837190
    128 rdf:type schema:CreativeWork
    129 https://doi.org/10.1109/cvpr.2017.243 schema:sameAs https://app.dimensions.ai/details/publication/pub.1095843442
    130 rdf:type schema:CreativeWork
    131 https://doi.org/10.1109/iccv.2015.169 schema:sameAs https://app.dimensions.ai/details/publication/pub.1095573598
    132 rdf:type schema:CreativeWork
    133 https://doi.org/10.1109/iccv.2017.155 schema:sameAs https://app.dimensions.ai/details/publication/pub.1100060121
    134 rdf:type schema:CreativeWork
    135 https://doi.org/10.1109/tpami.2016.2577031 schema:sameAs https://app.dimensions.ai/details/publication/pub.1061745117
    136 rdf:type schema:CreativeWork
    137 https://doi.org/10.1109/tpami.2016.2644615 schema:sameAs https://app.dimensions.ai/details/publication/pub.1061745216
    138 rdf:type schema:CreativeWork
    139 https://doi.org/10.1145/3005348 schema:sameAs https://app.dimensions.ai/details/publication/pub.1083756204
    140 rdf:type schema:CreativeWork
    141 https://doi.org/10.1177/0269215511431903 schema:sameAs https://app.dimensions.ai/details/publication/pub.1032913631
    142 rdf:type schema:CreativeWork
    143 https://doi.org/10.5244/c.28.88 schema:sameAs https://app.dimensions.ai/details/publication/pub.1099426768
    144 rdf:type schema:CreativeWork
    145 https://www.grid.ac/institutes/grid.417969.4 schema:alternateName Indian Institute of Technology Madras
    146 schema:name Department of Computer Science and Engineering, Robert Bosch Centre for Data Science and AI (RBC-DSAI), Indian Institute of Technology Madras, Chennai, India
    147 rdf:type schema:Organization
     




    Preview window. Press ESC to close (or click here)


    ...