Visualizing and Understanding Convolutional Networks View Full Text


Ontology type: schema:Chapter      Open Access: True


Chapter Info

DATE

2014

AUTHORS

Matthew D. Zeiler , Rob Fergus

ABSTRACT

Large Convolutional Network models have recently demonstrated impressive classification performance on the ImageNet benchmark Krizhevsky et al. [18]. However there is no clear understanding of why they perform so well, or how they might be improved. In this paper we explore both issues. We introduce a novel visualization technique that gives insight into the function of intermediate feature layers and the operation of the classifier. Used in a diagnostic role, these visualizations allow us to find model architectures that outperform Krizhevsky et al on the ImageNet classification benchmark. We also perform an ablation study to discover the performance contribution from different model layers. We show our ImageNet model generalizes well to other datasets: when the softmax classifier is retrained, it convincingly beats the current state-of-the-art results on Caltech-101 and Caltech-256 datasets. More... »

PAGES

818-833

Identifiers

URI

http://scigraph.springernature.com/pub.10.1007/978-3-319-10590-1_53

DOI

http://dx.doi.org/10.1007/978-3-319-10590-1_53

DIMENSIONS

https://app.dimensions.ai/details/publication/pub.1032233097


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/2008/0102", 
        "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
        "name": "Applied Mathematics", 
        "type": "DefinedTerm"
      }, 
      {
        "id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/01", 
        "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
        "name": "Mathematical Sciences", 
        "type": "DefinedTerm"
      }
    ], 
    "author": [
      {
        "affiliation": {
          "alternateName": "New York University", 
          "id": "https://www.grid.ac/institutes/grid.137628.9", 
          "name": [
            "Dept. of Computer Science, New York University, USA"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Zeiler", 
        "givenName": "Matthew D.", 
        "id": "sg:person.012065541527.33", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012065541527.33"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "New York University", 
          "id": "https://www.grid.ac/institutes/grid.137628.9", 
          "name": [
            "Dept. of Computer Science, New York University, USA"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Fergus", 
        "givenName": "Rob", 
        "id": "sg:person.011304642427.80", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.011304642427.80"
        ], 
        "type": "Person"
      }
    ], 
    "citation": [
      {
        "id": "https://doi.org/10.1162/neco.2006.18.7.1527", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1004707137"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1162/neco.1989.1.4.541", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1008345178"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1145/1390156.1390294", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1034603392"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1162/neco.2006.18.8.1868", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1052237072"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/tpami.2006.79", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1061743121"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/iccv.2009.5459469", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1093416695"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/cvpr.2014.222", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1094012327"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/cvpr.2011.5995347", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1094016389"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/cvpr.2013.91", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1094629530"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/cvpr.2013.91", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1094629530"
        ], 
        "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/cvpr.2009.5206757", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1095180230"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/iccv.2011.6126474", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1095432752"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/cvpr.2009.5206848", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1095689025"
        ], 
        "type": "CreativeWork"
      }, 
      {
        "id": "https://doi.org/10.1109/iccv.2011.6126554", 
        "sameAs": [
          "https://app.dimensions.ai/details/publication/pub.1095733216"
        ], 
        "type": "CreativeWork"
      }
    ], 
    "datePublished": "2014", 
    "datePublishedReg": "2014-01-01", 
    "description": "Large Convolutional Network models have recently demonstrated impressive classification performance on the ImageNet benchmark Krizhevsky et al. [18]. However there is no clear understanding of why they perform so well, or how they might be improved. In this paper we explore both issues. We introduce a novel visualization technique that gives insight into the function of intermediate feature layers and the operation of the classifier. Used in a diagnostic role, these visualizations allow us to find model architectures that outperform Krizhevsky et al on the ImageNet classification benchmark. We also perform an ablation study to discover the performance contribution from different model layers. We show our ImageNet model generalizes well to other datasets: when the softmax classifier is retrained, it convincingly beats the current state-of-the-art results on Caltech-101 and Caltech-256 datasets.", 
    "editor": [
      {
        "familyName": "Fleet", 
        "givenName": "David", 
        "type": "Person"
      }, 
      {
        "familyName": "Pajdla", 
        "givenName": "Tomas", 
        "type": "Person"
      }, 
      {
        "familyName": "Schiele", 
        "givenName": "Bernt", 
        "type": "Person"
      }, 
      {
        "familyName": "Tuytelaars", 
        "givenName": "Tinne", 
        "type": "Person"
      }
    ], 
    "genre": "chapter", 
    "id": "sg:pub.10.1007/978-3-319-10590-1_53", 
    "inLanguage": [
      "en"
    ], 
    "isAccessibleForFree": true, 
    "isPartOf": {
      "isbn": [
        "978-3-319-10589-5", 
        "978-3-319-10590-1"
      ], 
      "name": "Computer Vision \u2013 ECCV 2014", 
      "type": "Book"
    }, 
    "name": "Visualizing and Understanding Convolutional Networks", 
    "pagination": "818-833", 
    "productId": [
      {
        "name": "doi", 
        "type": "PropertyValue", 
        "value": [
          "10.1007/978-3-319-10590-1_53"
        ]
      }, 
      {
        "name": "readcube_id", 
        "type": "PropertyValue", 
        "value": [
          "998d8a3218ba1a5b2d72024c4ae9832b9e776dee889c0ca05b3490424ddbdaf7"
        ]
      }, 
      {
        "name": "dimensions_id", 
        "type": "PropertyValue", 
        "value": [
          "pub.1032233097"
        ]
      }
    ], 
    "publisher": {
      "location": "Cham", 
      "name": "Springer International Publishing", 
      "type": "Organisation"
    }, 
    "sameAs": [
      "https://doi.org/10.1007/978-3-319-10590-1_53", 
      "https://app.dimensions.ai/details/publication/pub.1032233097"
    ], 
    "sdDataset": "chapters", 
    "sdDatePublished": "2019-04-15T19:42", 
    "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/v90/0000000001_0000000264/records_8684_00000554.jsonl", 
    "type": "Chapter", 
    "url": "http://link.springer.com/10.1007/978-3-319-10590-1_53"
  }
]
 

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.10.1007/978-3-319-10590-1_53'

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

curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1007/978-3-319-10590-1_53'

Turtle is a human-readable linked data format.

curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/978-3-319-10590-1_53'

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

curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/978-3-319-10590-1_53'


 

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

129 TRIPLES      23 PREDICATES      41 URIs      20 LITERALS      8 BLANK NODES

Subject Predicate Object
1 sg:pub.10.1007/978-3-319-10590-1_53 schema:about anzsrc-for:01
2 anzsrc-for:0102
3 schema:author Nba3106be33f04ad79349feefa2426a55
4 schema:citation https://doi.org/10.1109/cvpr.2009.5206757
5 https://doi.org/10.1109/cvpr.2009.5206848
6 https://doi.org/10.1109/cvpr.2011.5995347
7 https://doi.org/10.1109/cvpr.2013.91
8 https://doi.org/10.1109/cvpr.2014.222
9 https://doi.org/10.1109/cvpr.2014.81
10 https://doi.org/10.1109/iccv.2009.5459469
11 https://doi.org/10.1109/iccv.2011.6126474
12 https://doi.org/10.1109/iccv.2011.6126554
13 https://doi.org/10.1109/tpami.2006.79
14 https://doi.org/10.1145/1390156.1390294
15 https://doi.org/10.1162/neco.1989.1.4.541
16 https://doi.org/10.1162/neco.2006.18.7.1527
17 https://doi.org/10.1162/neco.2006.18.8.1868
18 schema:datePublished 2014
19 schema:datePublishedReg 2014-01-01
20 schema:description Large Convolutional Network models have recently demonstrated impressive classification performance on the ImageNet benchmark Krizhevsky et al. [18]. However there is no clear understanding of why they perform so well, or how they might be improved. In this paper we explore both issues. We introduce a novel visualization technique that gives insight into the function of intermediate feature layers and the operation of the classifier. Used in a diagnostic role, these visualizations allow us to find model architectures that outperform Krizhevsky et al on the ImageNet classification benchmark. We also perform an ablation study to discover the performance contribution from different model layers. We show our ImageNet model generalizes well to other datasets: when the softmax classifier is retrained, it convincingly beats the current state-of-the-art results on Caltech-101 and Caltech-256 datasets.
21 schema:editor N45dac717767b4f4d8046bf21faf97772
22 schema:genre chapter
23 schema:inLanguage en
24 schema:isAccessibleForFree true
25 schema:isPartOf N635333099d4f493e83484e8625b8c74a
26 schema:name Visualizing and Understanding Convolutional Networks
27 schema:pagination 818-833
28 schema:productId N25183c7aafaa458fad256feed5fb72a6
29 N8a50032a927b4b5783f723b0b137ad96
30 Na67f7bb43adb4cb796a9fd2814257e1a
31 schema:publisher N3034089b1c2c4fc9b1fde776bf7deaf9
32 schema:sameAs https://app.dimensions.ai/details/publication/pub.1032233097
33 https://doi.org/10.1007/978-3-319-10590-1_53
34 schema:sdDatePublished 2019-04-15T19:42
35 schema:sdLicense https://scigraph.springernature.com/explorer/license/
36 schema:sdPublisher N36486ba7b2dc4159aa424503a3886121
37 schema:url http://link.springer.com/10.1007/978-3-319-10590-1_53
38 sgo:license sg:explorer/license/
39 sgo:sdDataset chapters
40 rdf:type schema:Chapter
41 N1fc4378e067244149850af50525e6fb9 rdf:first N4287cf70c7564c1482026148cac860e8
42 rdf:rest rdf:nil
43 N25183c7aafaa458fad256feed5fb72a6 schema:name readcube_id
44 schema:value 998d8a3218ba1a5b2d72024c4ae9832b9e776dee889c0ca05b3490424ddbdaf7
45 rdf:type schema:PropertyValue
46 N3034089b1c2c4fc9b1fde776bf7deaf9 schema:location Cham
47 schema:name Springer International Publishing
48 rdf:type schema:Organisation
49 N36486ba7b2dc4159aa424503a3886121 schema:name Springer Nature - SN SciGraph project
50 rdf:type schema:Organization
51 N3ef6c7a528fd44ad9b6c5f828ed3948b rdf:first sg:person.011304642427.80
52 rdf:rest rdf:nil
53 N422addd55a2f4988a2d25d7ba43cfaa2 rdf:first N789ef6c41de44c398151f4a47f26f19e
54 rdf:rest N1fc4378e067244149850af50525e6fb9
55 N4287cf70c7564c1482026148cac860e8 schema:familyName Tuytelaars
56 schema:givenName Tinne
57 rdf:type schema:Person
58 N45dac717767b4f4d8046bf21faf97772 rdf:first N4d0f680129a74b19af100ab9d91cd2c5
59 rdf:rest Naefe2c39574141049c2e720cb77b2bd1
60 N4d0f680129a74b19af100ab9d91cd2c5 schema:familyName Fleet
61 schema:givenName David
62 rdf:type schema:Person
63 N635333099d4f493e83484e8625b8c74a schema:isbn 978-3-319-10589-5
64 978-3-319-10590-1
65 schema:name Computer Vision – ECCV 2014
66 rdf:type schema:Book
67 N789ef6c41de44c398151f4a47f26f19e schema:familyName Schiele
68 schema:givenName Bernt
69 rdf:type schema:Person
70 N8a50032a927b4b5783f723b0b137ad96 schema:name dimensions_id
71 schema:value pub.1032233097
72 rdf:type schema:PropertyValue
73 Na67f7bb43adb4cb796a9fd2814257e1a schema:name doi
74 schema:value 10.1007/978-3-319-10590-1_53
75 rdf:type schema:PropertyValue
76 Naefe2c39574141049c2e720cb77b2bd1 rdf:first Ncaf9898ba61d4f5c89244f6c7bf4b735
77 rdf:rest N422addd55a2f4988a2d25d7ba43cfaa2
78 Nba3106be33f04ad79349feefa2426a55 rdf:first sg:person.012065541527.33
79 rdf:rest N3ef6c7a528fd44ad9b6c5f828ed3948b
80 Ncaf9898ba61d4f5c89244f6c7bf4b735 schema:familyName Pajdla
81 schema:givenName Tomas
82 rdf:type schema:Person
83 anzsrc-for:01 schema:inDefinedTermSet anzsrc-for:
84 schema:name Mathematical Sciences
85 rdf:type schema:DefinedTerm
86 anzsrc-for:0102 schema:inDefinedTermSet anzsrc-for:
87 schema:name Applied Mathematics
88 rdf:type schema:DefinedTerm
89 sg:person.011304642427.80 schema:affiliation https://www.grid.ac/institutes/grid.137628.9
90 schema:familyName Fergus
91 schema:givenName Rob
92 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.011304642427.80
93 rdf:type schema:Person
94 sg:person.012065541527.33 schema:affiliation https://www.grid.ac/institutes/grid.137628.9
95 schema:familyName Zeiler
96 schema:givenName Matthew D.
97 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012065541527.33
98 rdf:type schema:Person
99 https://doi.org/10.1109/cvpr.2009.5206757 schema:sameAs https://app.dimensions.ai/details/publication/pub.1095180230
100 rdf:type schema:CreativeWork
101 https://doi.org/10.1109/cvpr.2009.5206848 schema:sameAs https://app.dimensions.ai/details/publication/pub.1095689025
102 rdf:type schema:CreativeWork
103 https://doi.org/10.1109/cvpr.2011.5995347 schema:sameAs https://app.dimensions.ai/details/publication/pub.1094016389
104 rdf:type schema:CreativeWork
105 https://doi.org/10.1109/cvpr.2013.91 schema:sameAs https://app.dimensions.ai/details/publication/pub.1094629530
106 rdf:type schema:CreativeWork
107 https://doi.org/10.1109/cvpr.2014.222 schema:sameAs https://app.dimensions.ai/details/publication/pub.1094012327
108 rdf:type schema:CreativeWork
109 https://doi.org/10.1109/cvpr.2014.81 schema:sameAs https://app.dimensions.ai/details/publication/pub.1094727707
110 rdf:type schema:CreativeWork
111 https://doi.org/10.1109/iccv.2009.5459469 schema:sameAs https://app.dimensions.ai/details/publication/pub.1093416695
112 rdf:type schema:CreativeWork
113 https://doi.org/10.1109/iccv.2011.6126474 schema:sameAs https://app.dimensions.ai/details/publication/pub.1095432752
114 rdf:type schema:CreativeWork
115 https://doi.org/10.1109/iccv.2011.6126554 schema:sameAs https://app.dimensions.ai/details/publication/pub.1095733216
116 rdf:type schema:CreativeWork
117 https://doi.org/10.1109/tpami.2006.79 schema:sameAs https://app.dimensions.ai/details/publication/pub.1061743121
118 rdf:type schema:CreativeWork
119 https://doi.org/10.1145/1390156.1390294 schema:sameAs https://app.dimensions.ai/details/publication/pub.1034603392
120 rdf:type schema:CreativeWork
121 https://doi.org/10.1162/neco.1989.1.4.541 schema:sameAs https://app.dimensions.ai/details/publication/pub.1008345178
122 rdf:type schema:CreativeWork
123 https://doi.org/10.1162/neco.2006.18.7.1527 schema:sameAs https://app.dimensions.ai/details/publication/pub.1004707137
124 rdf:type schema:CreativeWork
125 https://doi.org/10.1162/neco.2006.18.8.1868 schema:sameAs https://app.dimensions.ai/details/publication/pub.1052237072
126 rdf:type schema:CreativeWork
127 https://www.grid.ac/institutes/grid.137628.9 schema:alternateName New York University
128 schema:name Dept. of Computer Science, New York University, USA
129 rdf:type schema:Organization
 




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


...