Descriptor Learning for Efficient Retrieval View Full Text


Ontology type: schema:Chapter      Open Access: True


Chapter Info

DATE

2010

AUTHORS

James Philbin , Michael Isard , Josef Sivic , Andrew Zisserman

ABSTRACT

Many visual search and matching systems represent images using sparse sets of “visual words”: descriptors that have been quantized by assignment to the best-matching symbol in a discrete vocabulary. Errors in this quantization procedure propagate throughout the rest of the system, either harming performance or requiring correction using additional storage or processing. This paper aims to reduce these quantization errors at source, by learning a projection from descriptor space to a new Euclidean space in which standard clustering techniques are more likely to assign matching descriptors to the same cluster, and non-matching descriptors to different clusters.To achieve this, we learn a non-linear transformation model by minimizing a novel margin-based cost function, which aims to separate matching descriptors from two classes of non-matching descriptors. Training data is generated automatically by leveraging geometric consistency. Scalable, stochastic gradient methods are used for the optimization.For the case of particular object retrieval, we demonstrate impressive gains in performance on a ground truth dataset: our learnt 32-D descriptor without spatial re-ranking outperforms a baseline method using 128-D SIFT descriptors with spatial re-ranking. More... »

PAGES

677-691

Identifiers

URI

http://scigraph.springernature.com/pub.10.1007/978-3-642-15558-1_49

DOI

http://dx.doi.org/10.1007/978-3-642-15558-1_49

DIMENSIONS

https://app.dimensions.ai/details/publication/pub.1045882552


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": "Visual Geometry Group, Department of Engineering Science, University of Oxford", 
          "id": "http://www.grid.ac/institutes/grid.4991.5", 
          "name": [
            "Visual Geometry Group, Department of Engineering Science, University of Oxford"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Philbin", 
        "givenName": "James", 
        "id": "sg:person.012204071073.98", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012204071073.98"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "Microsoft Research, Silicon Valley", 
          "id": "http://www.grid.ac/institutes/grid.419815.0", 
          "name": [
            "Microsoft Research, Silicon Valley"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Isard", 
        "givenName": "Michael", 
        "id": "sg:person.07376125500.79", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.07376125500.79"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "WILLOW, Laboratoire d\u2019Informatique de l\u2019Ecole Normale Superieure, INRIA, Paris", 
          "id": "http://www.grid.ac/institutes/grid.5328.c", 
          "name": [
            "WILLOW, Laboratoire d\u2019Informatique de l\u2019Ecole Normale Superieure, INRIA, Paris"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Sivic", 
        "givenName": "Josef", 
        "id": "sg:person.0757630654.27", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.0757630654.27"
        ], 
        "type": "Person"
      }, 
      {
        "affiliation": {
          "alternateName": "Visual Geometry Group, Department of Engineering Science, University of Oxford", 
          "id": "http://www.grid.ac/institutes/grid.4991.5", 
          "name": [
            "Visual Geometry Group, Department of Engineering Science, University of Oxford"
          ], 
          "type": "Organization"
        }, 
        "familyName": "Zisserman", 
        "givenName": "Andrew", 
        "id": "sg:person.012270111307.09", 
        "sameAs": [
          "https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012270111307.09"
        ], 
        "type": "Person"
      }
    ], 
    "datePublished": "2010", 
    "datePublishedReg": "2010-01-01", 
    "description": "Many visual search and matching systems represent images using sparse sets of \u201cvisual words\u201d: descriptors that have been quantized by assignment to the best-matching symbol in a discrete vocabulary. Errors in this quantization procedure propagate throughout the rest of the system, either harming performance or requiring correction using additional storage or processing. This paper aims to reduce these quantization errors at source, by learning a projection from descriptor space to a new Euclidean space in which standard clustering techniques are more likely to assign matching descriptors to the same cluster, and non-matching descriptors to different clusters.To achieve this, we learn a non-linear transformation model by minimizing a novel margin-based cost function, which aims to separate matching descriptors from two classes of non-matching descriptors. Training data is generated automatically by leveraging geometric consistency. Scalable, stochastic gradient methods are used for the optimization.For the case of particular object retrieval, we demonstrate impressive gains in performance on a ground truth dataset: our learnt 32-D descriptor without spatial re-ranking outperforms a baseline method using 128-D SIFT descriptors with spatial re-ranking.", 
    "editor": [
      {
        "familyName": "Daniilidis", 
        "givenName": "Kostas", 
        "type": "Person"
      }, 
      {
        "familyName": "Maragos", 
        "givenName": "Petros", 
        "type": "Person"
      }, 
      {
        "familyName": "Paragios", 
        "givenName": "Nikos", 
        "type": "Person"
      }
    ], 
    "genre": "chapter", 
    "id": "sg:pub.10.1007/978-3-642-15558-1_49", 
    "inLanguage": "en", 
    "isAccessibleForFree": true, 
    "isPartOf": {
      "isbn": [
        "978-3-642-15557-4", 
        "978-3-642-15558-1"
      ], 
      "name": "Computer Vision \u2013 ECCV 2010", 
      "type": "Book"
    }, 
    "keywords": [
      "particular object retrieval", 
      "standard clustering techniques", 
      "ground truth dataset", 
      "stochastic gradient method", 
      "discrete vocabulary", 
      "descriptor learning", 
      "efficient retrieval", 
      "object retrieval", 
      "visual words", 
      "clustering techniques", 
      "SIFT descriptors", 
      "training data", 
      "geometric consistency", 
      "truth dataset", 
      "baseline methods", 
      "sparse set", 
      "quantization error", 
      "descriptor space", 
      "Euclidean space", 
      "same cluster", 
      "gradient method", 
      "new Euclidean space", 
      "cost function", 
      "additional storage", 
      "descriptors", 
      "different clusters", 
      "retrieval", 
      "transformation model", 
      "visual search", 
      "impressive gains", 
      "space", 
      "datasets", 
      "performance", 
      "error", 
      "system", 
      "images", 
      "optimization", 
      "learning", 
      "processing", 
      "vocabulary", 
      "clusters", 
      "set", 
      "search", 
      "class", 
      "method", 
      "symbols", 
      "technique", 
      "assignment", 
      "storage", 
      "words", 
      "model", 
      "consistency", 
      "projections", 
      "function", 
      "data", 
      "gain", 
      "cases", 
      "correction", 
      "source", 
      "propagates", 
      "rest", 
      "paper", 
      "quantization procedure propagate", 
      "procedure propagate", 
      "non-matching descriptors", 
      "non-linear transformation model", 
      "novel margin-based cost function", 
      "margin-based cost function", 
      "matching descriptors"
    ], 
    "name": "Descriptor Learning for Efficient Retrieval", 
    "pagination": "677-691", 
    "productId": [
      {
        "name": "dimensions_id", 
        "type": "PropertyValue", 
        "value": [
          "pub.1045882552"
        ]
      }, 
      {
        "name": "doi", 
        "type": "PropertyValue", 
        "value": [
          "10.1007/978-3-642-15558-1_49"
        ]
      }
    ], 
    "publisher": {
      "name": "Springer Nature", 
      "type": "Organisation"
    }, 
    "sameAs": [
      "https://doi.org/10.1007/978-3-642-15558-1_49", 
      "https://app.dimensions.ai/details/publication/pub.1045882552"
    ], 
    "sdDataset": "chapters", 
    "sdDatePublished": "2021-11-01T18:58", 
    "sdLicense": "https://scigraph.springernature.com/explorer/license/", 
    "sdPublisher": {
      "name": "Springer Nature - SN SciGraph project", 
      "type": "Organization"
    }, 
    "sdSource": "s3://com-springernature-scigraph/baseset/20211101/entities/gbq_results/chapter/chapter_378.jsonl", 
    "type": "Chapter", 
    "url": "https://doi.org/10.1007/978-3-642-15558-1_49"
  }
]
 

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.10.1007/978-3-642-15558-1_49'

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

curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1007/978-3-642-15558-1_49'

Turtle is a human-readable linked data format.

curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/978-3-642-15558-1_49'

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

curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/978-3-642-15558-1_49'


 

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

166 TRIPLES      23 PREDICATES      95 URIs      88 LITERALS      7 BLANK NODES

Subject Predicate Object
1 sg:pub.10.1007/978-3-642-15558-1_49 schema:about anzsrc-for:08
2 anzsrc-for:0801
3 schema:author N12f8d1acdb0c482a8969cecbd6c86bfc
4 schema:datePublished 2010
5 schema:datePublishedReg 2010-01-01
6 schema:description Many visual search and matching systems represent images using sparse sets of “visual words”: descriptors that have been quantized by assignment to the best-matching symbol in a discrete vocabulary. Errors in this quantization procedure propagate throughout the rest of the system, either harming performance or requiring correction using additional storage or processing. This paper aims to reduce these quantization errors at source, by learning a projection from descriptor space to a new Euclidean space in which standard clustering techniques are more likely to assign matching descriptors to the same cluster, and non-matching descriptors to different clusters.To achieve this, we learn a non-linear transformation model by minimizing a novel margin-based cost function, which aims to separate matching descriptors from two classes of non-matching descriptors. Training data is generated automatically by leveraging geometric consistency. Scalable, stochastic gradient methods are used for the optimization.For the case of particular object retrieval, we demonstrate impressive gains in performance on a ground truth dataset: our learnt 32-D descriptor without spatial re-ranking outperforms a baseline method using 128-D SIFT descriptors with spatial re-ranking.
7 schema:editor Nf55c67bb263f4fed952f12f7fff6f2dc
8 schema:genre chapter
9 schema:inLanguage en
10 schema:isAccessibleForFree true
11 schema:isPartOf Ne029207ea18445b3a8a4d790b4361106
12 schema:keywords Euclidean space
13 SIFT descriptors
14 additional storage
15 assignment
16 baseline methods
17 cases
18 class
19 clustering techniques
20 clusters
21 consistency
22 correction
23 cost function
24 data
25 datasets
26 descriptor learning
27 descriptor space
28 descriptors
29 different clusters
30 discrete vocabulary
31 efficient retrieval
32 error
33 function
34 gain
35 geometric consistency
36 gradient method
37 ground truth dataset
38 images
39 impressive gains
40 learning
41 margin-based cost function
42 matching descriptors
43 method
44 model
45 new Euclidean space
46 non-linear transformation model
47 non-matching descriptors
48 novel margin-based cost function
49 object retrieval
50 optimization
51 paper
52 particular object retrieval
53 performance
54 procedure propagate
55 processing
56 projections
57 propagates
58 quantization error
59 quantization procedure propagate
60 rest
61 retrieval
62 same cluster
63 search
64 set
65 source
66 space
67 sparse set
68 standard clustering techniques
69 stochastic gradient method
70 storage
71 symbols
72 system
73 technique
74 training data
75 transformation model
76 truth dataset
77 visual search
78 visual words
79 vocabulary
80 words
81 schema:name Descriptor Learning for Efficient Retrieval
82 schema:pagination 677-691
83 schema:productId N5c37d7871c6949b6943d21d7d525b403
84 N743f9f4c479c4a668409345bec66345e
85 schema:publisher N9ebb9ad5c4ea4f8a9438bdda391568cd
86 schema:sameAs https://app.dimensions.ai/details/publication/pub.1045882552
87 https://doi.org/10.1007/978-3-642-15558-1_49
88 schema:sdDatePublished 2021-11-01T18:58
89 schema:sdLicense https://scigraph.springernature.com/explorer/license/
90 schema:sdPublisher N6ece62c55aec4edc968fd62713a58179
91 schema:url https://doi.org/10.1007/978-3-642-15558-1_49
92 sgo:license sg:explorer/license/
93 sgo:sdDataset chapters
94 rdf:type schema:Chapter
95 N12f8d1acdb0c482a8969cecbd6c86bfc rdf:first sg:person.012204071073.98
96 rdf:rest N84a49318d30a4f76894b61ccc264ebfe
97 N3352b9f48e9b4d51a327e53d821e2290 rdf:first sg:person.012270111307.09
98 rdf:rest rdf:nil
99 N5c37d7871c6949b6943d21d7d525b403 schema:name dimensions_id
100 schema:value pub.1045882552
101 rdf:type schema:PropertyValue
102 N6ece62c55aec4edc968fd62713a58179 schema:name Springer Nature - SN SciGraph project
103 rdf:type schema:Organization
104 N743f9f4c479c4a668409345bec66345e schema:name doi
105 schema:value 10.1007/978-3-642-15558-1_49
106 rdf:type schema:PropertyValue
107 N84a49318d30a4f76894b61ccc264ebfe rdf:first sg:person.07376125500.79
108 rdf:rest Neae4d5198947425ea50ac6693419bee1
109 N9ebb9ad5c4ea4f8a9438bdda391568cd schema:name Springer Nature
110 rdf:type schema:Organisation
111 N9fa1af3989254b60ac152c80cffbb5c6 rdf:first Nae970da08f3a487ca7aecc200faf5b99
112 rdf:rest rdf:nil
113 Na23af30808df46dd981be7d3a1137162 schema:familyName Daniilidis
114 schema:givenName Kostas
115 rdf:type schema:Person
116 Nae970da08f3a487ca7aecc200faf5b99 schema:familyName Paragios
117 schema:givenName Nikos
118 rdf:type schema:Person
119 Nb3e0d6c98d1344ca848c0ffde53ad143 rdf:first Ndc4e652f73b64becbfa14283ac3437fa
120 rdf:rest N9fa1af3989254b60ac152c80cffbb5c6
121 Ndc4e652f73b64becbfa14283ac3437fa schema:familyName Maragos
122 schema:givenName Petros
123 rdf:type schema:Person
124 Ne029207ea18445b3a8a4d790b4361106 schema:isbn 978-3-642-15557-4
125 978-3-642-15558-1
126 schema:name Computer Vision – ECCV 2010
127 rdf:type schema:Book
128 Neae4d5198947425ea50ac6693419bee1 rdf:first sg:person.0757630654.27
129 rdf:rest N3352b9f48e9b4d51a327e53d821e2290
130 Nf55c67bb263f4fed952f12f7fff6f2dc rdf:first Na23af30808df46dd981be7d3a1137162
131 rdf:rest Nb3e0d6c98d1344ca848c0ffde53ad143
132 anzsrc-for:08 schema:inDefinedTermSet anzsrc-for:
133 schema:name Information and Computing Sciences
134 rdf:type schema:DefinedTerm
135 anzsrc-for:0801 schema:inDefinedTermSet anzsrc-for:
136 schema:name Artificial Intelligence and Image Processing
137 rdf:type schema:DefinedTerm
138 sg:person.012204071073.98 schema:affiliation grid-institutes:grid.4991.5
139 schema:familyName Philbin
140 schema:givenName James
141 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012204071073.98
142 rdf:type schema:Person
143 sg:person.012270111307.09 schema:affiliation grid-institutes:grid.4991.5
144 schema:familyName Zisserman
145 schema:givenName Andrew
146 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012270111307.09
147 rdf:type schema:Person
148 sg:person.07376125500.79 schema:affiliation grid-institutes:grid.419815.0
149 schema:familyName Isard
150 schema:givenName Michael
151 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.07376125500.79
152 rdf:type schema:Person
153 sg:person.0757630654.27 schema:affiliation grid-institutes:grid.5328.c
154 schema:familyName Sivic
155 schema:givenName Josef
156 schema:sameAs https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.0757630654.27
157 rdf:type schema:Person
158 grid-institutes:grid.419815.0 schema:alternateName Microsoft Research, Silicon Valley
159 schema:name Microsoft Research, Silicon Valley
160 rdf:type schema:Organization
161 grid-institutes:grid.4991.5 schema:alternateName Visual Geometry Group, Department of Engineering Science, University of Oxford
162 schema:name Visual Geometry Group, Department of Engineering Science, University of Oxford
163 rdf:type schema:Organization
164 grid-institutes:grid.5328.c schema:alternateName WILLOW, Laboratoire d’Informatique de l’Ecole Normale Superieure, INRIA, Paris
165 schema:name WILLOW, Laboratoire d’Informatique de l’Ecole Normale Superieure, INRIA, Paris
166 rdf:type schema:Organization
 




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


...