Long-Tailed Recognition Using Class-Balanced Experts View Full Text


Ontology type: schema:Chapter      Open Access: True


Chapter Info

DATE

2021-03-17

AUTHORS

Saurabh Sharma , Ning Yu , Mario Fritz , Bernt Schiele

ABSTRACT

Deep learning enables impressive performance in image recognition using large-scale artificially-balanced datasets. However, real-world datasets exhibit highly class-imbalanced distributions, yielding two main challenges: relative imbalance amongst the classes and data scarcity for mediumshot or fewshot classes. In this work, we address the problem of long-tailed recognition wherein the training set is highly imbalanced and the test set is kept balanced. Differently from existing paradigms relying on data-resampling, cost-sensitive learning, online hard example mining, loss objective reshaping, and/or memory-based modeling, we propose an ensemble of class-balanced experts that combines the strength of diverse classifiers. Our ensemble of class-balanced experts reaches results close to state-of-the-art and an extended ensemble establishes a new state-of-the-art on two benchmarks for long-tailed recognition. We conduct extensive experiments to analyse the performance of the ensembles, and discover that in modern large-scale datasets, relative imbalance is a harder problem than data scarcity. The training and evaluation code is available at https://github.com/ssfootball04/class-balanced-experts. More... »

PAGES

86-100

Identifiers

URI

http://scigraph.springernature.com/pub.10.1007/978-3-030-71278-5_7

DOI

http://dx.doi.org/10.1007/978-3-030-71278-5_7

DIMENSIONS

https://app.dimensions.ai/details/publication/pub.1136442854


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/2008/08", 
        "inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/", 
        "name": "Information and Computing Sciences", 
        "type": "DefinedTerm"
      }, 
      {
        "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"
      }
    ], 
    "author": [
      {
        "affiliation": {
          "alternateName": "Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbr\u00fccken, Germany", 
          "id": "http://www.grid.ac/institutes/grid.419528.3", 
          "name": [
            "Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbr\u00fccken, Germany"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Sharma", 
        "givenName": "Saurabh", 
        "id": "sg:person.016335003431.12", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.016335003431.12"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "University of Maryland, College Park, USA", 
          "id": "http://www.grid.ac/institutes/grid.164295.d", 
          "name": [
            "Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbr\u00fccken, Germany", 
            "University of Maryland, College Park, USA"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Yu", 
        "givenName": "Ning", 
        "id": "sg:person.016547261025.90", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.016547261025.90"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "CISPA Helmholtz Center for Information Security, Saarland Informatics Campus, Saarbr\u00fccken, Germany", 
          "id": "http://www.grid.ac/institutes/grid.507511.7", 
          "name": [
            "CISPA Helmholtz Center for Information Security, Saarland Informatics Campus, Saarbr\u00fccken, Germany"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Fritz", 
        "givenName": "Mario", 
        "id": "sg:person.013361072755.17", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013361072755.17"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbr\u00fccken, Germany", 
          "id": "http://www.grid.ac/institutes/grid.419528.3", 
          "name": [
            "Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbr\u00fccken, Germany"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Schiele", 
        "givenName": "Bernt", 
        "id": "sg:person.01174260421.90", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01174260421.90"
        ], 
        "type": "Person"
      }
    ], 
    "datePublished": "2021-03-17", 
    "datePublishedReg": "2021-03-17", 
    "description": "Deep learning enables impressive performance in image recognition using large-scale artificially-balanced datasets. However, real-world datasets exhibit highly class-imbalanced distributions, yielding two main challenges: relative imbalance amongst the classes and data scarcity for mediumshot or fewshot classes. In this work, we address the problem of long-tailed recognition wherein the training set is highly imbalanced and the test set is kept balanced. Differently from existing paradigms relying on data-resampling, cost-sensitive learning, online hard example mining, loss objective reshaping, and/or memory-based modeling, we propose an ensemble of class-balanced experts that combines the strength of diverse classifiers. Our ensemble of class-balanced experts reaches results close to state-of-the-art and an extended ensemble establishes a new state-of-the-art on two benchmarks for long-tailed recognition. We conduct extensive experiments to analyse the performance of the ensembles, and discover that in modern large-scale datasets, relative imbalance is a harder problem than data scarcity. The training and evaluation code is available at https://github.com/ssfootball04/class-balanced-experts.", 
    "editor": [
      {
        "familyName": "Akata", 
        "givenName": "Zeynep", 
        "type": "Person"
      }, 
      {
        "familyName": "Geiger", 
        "givenName": "Andreas", 
        "type": "Person"
      }, 
      {
        "familyName": "Sattler", 
        "givenName": "Torsten", 
        "type": "Person"
      }
    ], 
    "genre": "chapter", 
    "id": "sg:pub.10.1007/978-3-030-71278-5_7", 
    "isAccessibleForFree": true, 
    "isPartOf": {
      "isbn": [
        "978-3-030-71277-8", 
        "978-3-030-71278-5"
      ], 
      "name": "Pattern Recognition", 
      "type": "Book"
    }, 
    "keywords": [
      "online hard example mining", 
      "class-imbalanced distribution", 
      "hard example mining", 
      "real-world datasets", 
      "cost-sensitive learning", 
      "large-scale datasets", 
      "data scarcity", 
      "image recognition", 
      "deep learning", 
      "Extensive experiments", 
      "diverse classifiers", 
      "hard problem", 
      "impressive performance", 
      "training set", 
      "test set", 
      "dataset", 
      "main challenges", 
      "evaluation code", 
      "learning", 
      "recognition", 
      "experts", 
      "new state", 
      "classifier", 
      "mining", 
      "set", 
      "ensemble", 
      "art", 
      "performance", 
      "benchmarks", 
      "code", 
      "paradigm", 
      "challenges", 
      "modeling", 
      "class", 
      "training", 
      "work", 
      "state", 
      "experiments", 
      "scarcity", 
      "extended ensemble", 
      "imbalance", 
      "results", 
      "distribution", 
      "reshaping", 
      "strength", 
      "relative imbalance", 
      "problem"
    ], 
    "name": "Long-Tailed Recognition Using Class-Balanced Experts", 
    "pagination": "86-100", 
    "productId": [
      {
        "name": "dimensions_id", 
        "type": "PropertyValue", 
        "value": [
          "pub.1136442854"
        ]
      }, 
      {
        "name": "doi", 
        "type": "PropertyValue", 
        "value": [
          "10.1007/978-3-030-71278-5_7"
        ]
      }
    ], 
    "publisher": {
      "name": "Springer Nature", 
      "type": "Organisation"
    }, 
    "sameAs": [
      "https://doi.org/10.1007/978-3-030-71278-5_7", 
      "https://app.dimensions.ai/details/publication/pub.1136442854"
    ], 
    "sdDataset": "chapters", 
    "sdDatePublished": "2022-11-24T21:11", 
    "sdLicense": "https://scigraph.springernature.com/explorer/license/", 
    "sdPublisher": {
      "name": "Springer Nature - SN SciGraph project", 
      "type": "Organization"
    }, 
    "sdSource": "s3://com-springernature-scigraph/baseset/20221124/entities/gbq_results/chapter/chapter_130.jsonl", 
    "type": "Chapter", 
    "url": "https://doi.org/10.1007/978-3-030-71278-5_7"
  }
]
 

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.10.1007/978-3-030-71278-5_7'

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

curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-71278-5_7'

Turtle is a human-readable linked data format.

curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-71278-5_7'

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

curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-71278-5_7'


 

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

144 TRIPLES      22 PREDICATES      71 URIs      64 LITERALS      7 BLANK NODES

Subject Predicate Object
1 sg:pub.10.1007/978-3-030-71278-5_7 schema:about anzsrc-for:08
2 anzsrc-for:0801
3 schema:author Na1937d1c53564935af89c3365463759c
4 schema:datePublished 2021-03-17
5 schema:datePublishedReg 2021-03-17
6 schema:description Deep learning enables impressive performance in image recognition using large-scale artificially-balanced datasets. However, real-world datasets exhibit highly class-imbalanced distributions, yielding two main challenges: relative imbalance amongst the classes and data scarcity for mediumshot or fewshot classes. In this work, we address the problem of long-tailed recognition wherein the training set is highly imbalanced and the test set is kept balanced. Differently from existing paradigms relying on data-resampling, cost-sensitive learning, online hard example mining, loss objective reshaping, and/or memory-based modeling, we propose an ensemble of class-balanced experts that combines the strength of diverse classifiers. Our ensemble of class-balanced experts reaches results close to state-of-the-art and an extended ensemble establishes a new state-of-the-art on two benchmarks for long-tailed recognition. We conduct extensive experiments to analyse the performance of the ensembles, and discover that in modern large-scale datasets, relative imbalance is a harder problem than data scarcity. The training and evaluation code is available at https://github.com/ssfootball04/class-balanced-experts.
7 schema:editor N02ee4644e5be4f6caa54ea56f5ed7fef
8 schema:genre chapter
9 schema:isAccessibleForFree true
10 schema:isPartOf Nda293f84b2e74770a8c1a31995a1b898
11 schema:keywords Extensive experiments
12 art
13 benchmarks
14 challenges
15 class
16 class-imbalanced distribution
17 classifier
18 code
19 cost-sensitive learning
20 data scarcity
21 dataset
22 deep learning
23 distribution
24 diverse classifiers
25 ensemble
26 evaluation code
27 experiments
28 experts
29 extended ensemble
30 hard example mining
31 hard problem
32 image recognition
33 imbalance
34 impressive performance
35 large-scale datasets
36 learning
37 main challenges
38 mining
39 modeling
40 new state
41 online hard example mining
42 paradigm
43 performance
44 problem
45 real-world datasets
46 recognition
47 relative imbalance
48 reshaping
49 results
50 scarcity
51 set
52 state
53 strength
54 test set
55 training
56 training set
57 work
58 schema:name Long-Tailed Recognition Using Class-Balanced Experts
59 schema:pagination 86-100
60 schema:productId N134d79ce3ea844fc8f8e91ca2aa88da6
61 N3274ec5ac4ed422099dd3bad7662e2e2
62 schema:publisher N27cc0390717346cc9cdcf0d8402882ae
63 schema:sameAs https://app.dimensions.ai/details/publication/pub.1136442854
64 https://doi.org/10.1007/978-3-030-71278-5_7
65 schema:sdDatePublished 2022-11-24T21:11
66 schema:sdLicense https://scigraph.springernature.com/explorer/license/
67 schema:sdPublisher N6e7f5505072d4128b838793bca4e7f79
68 schema:url https://doi.org/10.1007/978-3-030-71278-5_7
69 sgo:license sg:explorer/license/
70 sgo:sdDataset chapters
71 rdf:type schema:Chapter
72 N01586d03aa2b4047ba13cdfa2736a5ed schema:familyName Geiger
73 schema:givenName Andreas
74 rdf:type schema:Person
75 N02ee4644e5be4f6caa54ea56f5ed7fef rdf:first N45e0dc7fef9d4856aa01a38c82d45d1e
76 rdf:rest N665d7b85b75a4023a2c3b888bc180b6a
77 N134d79ce3ea844fc8f8e91ca2aa88da6 schema:name dimensions_id
78 schema:value pub.1136442854
79 rdf:type schema:PropertyValue
80 N2397daf254b2487b95f59e787b08ea73 rdf:first Nefc3ecbea1e649b49cbb1962dd72fbc6
81 rdf:rest rdf:nil
82 N27cc0390717346cc9cdcf0d8402882ae schema:name Springer Nature
83 rdf:type schema:Organisation
84 N3274ec5ac4ed422099dd3bad7662e2e2 schema:name doi
85 schema:value 10.1007/978-3-030-71278-5_7
86 rdf:type schema:PropertyValue
87 N45e0dc7fef9d4856aa01a38c82d45d1e schema:familyName Akata
88 schema:givenName Zeynep
89 rdf:type schema:Person
90 N665d7b85b75a4023a2c3b888bc180b6a rdf:first N01586d03aa2b4047ba13cdfa2736a5ed
91 rdf:rest N2397daf254b2487b95f59e787b08ea73
92 N6b88dcf7c89f473cb9e3185c610e8c88 rdf:first sg:person.01174260421.90
93 rdf:rest rdf:nil
94 N6e7f5505072d4128b838793bca4e7f79 schema:name Springer Nature - SN SciGraph project
95 rdf:type schema:Organization
96 N8b61b50c458b40d0a0b208faaa8edf68 rdf:first sg:person.013361072755.17
97 rdf:rest N6b88dcf7c89f473cb9e3185c610e8c88
98 Na1937d1c53564935af89c3365463759c rdf:first sg:person.016335003431.12
99 rdf:rest Ne4fe43fefd834e6d9e03318941c21de5
100 Nda293f84b2e74770a8c1a31995a1b898 schema:isbn 978-3-030-71277-8
101 978-3-030-71278-5
102 schema:name Pattern Recognition
103 rdf:type schema:Book
104 Ne4fe43fefd834e6d9e03318941c21de5 rdf:first sg:person.016547261025.90
105 rdf:rest N8b61b50c458b40d0a0b208faaa8edf68
106 Nefc3ecbea1e649b49cbb1962dd72fbc6 schema:familyName Sattler
107 schema:givenName Torsten
108 rdf:type schema:Person
109 anzsrc-for:08 schema:inDefinedTermSet anzsrc-for:
110 schema:name Information and Computing Sciences
111 rdf:type schema:DefinedTerm
112 anzsrc-for:0801 schema:inDefinedTermSet anzsrc-for:
113 schema:name Artificial Intelligence and Image Processing
114 rdf:type schema:DefinedTerm
115 sg:person.01174260421.90 schema:affiliation grid-institutes:grid.419528.3
116 schema:familyName Schiele
117 schema:givenName Bernt
118 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01174260421.90
119 rdf:type schema:Person
120 sg:person.013361072755.17 schema:affiliation grid-institutes:grid.507511.7
121 schema:familyName Fritz
122 schema:givenName Mario
123 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013361072755.17
124 rdf:type schema:Person
125 sg:person.016335003431.12 schema:affiliation grid-institutes:grid.419528.3
126 schema:familyName Sharma
127 schema:givenName Saurabh
128 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.016335003431.12
129 rdf:type schema:Person
130 sg:person.016547261025.90 schema:affiliation grid-institutes:grid.164295.d
131 schema:familyName Yu
132 schema:givenName Ning
133 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.016547261025.90
134 rdf:type schema:Person
135 grid-institutes:grid.164295.d schema:alternateName University of Maryland, College Park, USA
136 schema:name Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbrücken, Germany
137 University of Maryland, College Park, USA
138 rdf:type schema:Organization
139 grid-institutes:grid.419528.3 schema:alternateName Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbrücken, Germany
140 schema:name Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbrücken, Germany
141 rdf:type schema:Organization
142 grid-institutes:grid.507511.7 schema:alternateName CISPA Helmholtz Center for Information Security, Saarland Informatics Campus, Saarbrücken, Germany
143 schema:name CISPA Helmholtz Center for Information Security, Saarland Informatics Campus, Saarbrücken, Germany
144 rdf:type schema:Organization
 




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


...