The generalized 1M image-caption corpus

Generalized captions for all 1M images you can download here (file in format: flickr_id captions):



  • 1M images: Visual (Caption Corpus Ngrams)
  • 1M images: Visual (Google Ngrams)

    Original captions for all 1M images you can download here:



  • 1M image-caption corpus

  • questions or suggestions are welcome! contact point:


    Polina Kuznetsova, Email:

    Selected Examples


    Orig Big elm tree over the house is no their anymore .
    Ngram-Only (Caption Corpus Ngrams) Tree house .
    Saliency (Caption Corpus Ngrams) Tree over the house .
    Visual (Caption Corpus Ngrams) Tree over the house .
    Ngram-Only (Google Ngrams) Tree house .
    Saliency (Google Ngrams) Big elm tree over the house .
    Visual (Google Ngrams) Elm tree over the house .

    Orig The fish market in Pondicheri2 .
    Ngram-Only (Caption Corpus Ngrams) Fish market Pondicheri2 .
    Saliency (Caption Corpus Ngrams) The fish market .
    Visual (Caption Corpus Ngrams) The fish market .
    Ngram-Only (Google Ngrams) Fish market .
    Saliency (Google Ngrams) The fish market .
    Visual (Google Ngrams) The fish market .

    Orig Huge wall of glass at the Conference Centre in Yohohama Japan .
    Ngram-Only (Caption Corpus Ngrams) Wall of glass .
    Saliency (Caption Corpus Ngrams) Huge wall of glass .
    Visual (Caption Corpus Ngrams) Wall of glass .
    Ngram-Only (Google Ngrams) Wall of glass .
    Saliency (Google Ngrams) Huge wall of glass .
    Visual (Google Ngrams) Wall of glass at the Centre .

    Orig Pretty old building with no sign in New Haven , CT .
    Ngram-Only (Caption Corpus Ngrams) Building sign in .
    Saliency (Caption Corpus Ngrams) Old building .
    Visual (Caption Corpus Ngrams) Old building .
    Ngram-Only (Google Ngrams) Building sign in .
    Saliency (Google Ngrams) Old building with no sign .
    Visual (Google Ngrams) Building with no sign .

    Orig Random street sign in Blasewitz .
    Ngram-Only (Caption Corpus Ngrams) Street sign in .
    Saliency (Caption Corpus Ngrams) Street sign .
    Visual (Caption Corpus Ngrams) Street sign .
    Ngram-Only (Google Ngrams) Street sign .
    Saliency (Google Ngrams) Street sign .
    Visual (Google Ngrams) Street sign .

    Orig My footprint in a sand box .
    Ngram-Only (Caption Corpus Ngrams) Sand box .
    Saliency (Caption Corpus Ngrams) A sand box .
    Visual (Caption Corpus Ngrams) A sand box .
    Ngram-Only (Google Ngrams) Sand box .
    Saliency (Google Ngrams) My footprint in a sand box .
    Visual (Google Ngrams) My footprint in a sand box .

    Orig Beautiful red leaves in a back street of Freiburg .
    Ngram-Only (Caption Corpus Ngrams) Red leaves in a back street .
    Saliency (Caption Corpus Ngrams) Red leaves in a back street .
    Visual (Caption Corpus Ngrams) Red leaves in a back street .
    Ngram-Only (Google Ngrams) Red leaves in a back street .
    Saliency (Google Ngrams) Beautiful red leaves in a back street .
    Visual (Google Ngrams) Red leaves in a back street .


    More examples (1K images) can be found here >>

    Paper

    Generalizing Image Captions for Image-Text Parallel Corpus.
    Polina Kuznetsova, Vicente Ordonez, Alexander Berg, Tamara Berg and Yejin Choi.

    Association for Computational Linguistics (ACL), short, 2013

    Citation

    @InProceedings{kuznetsova-EtAl:2013:ACL2013,
      author    = {Kuznetsova, Polina  and  Ordonez, Vicente  and  Berg, Alexander  and  Berg, Tamara  and  Choi, Yejin},
      title     = {Generalizing Image Captions for Image-Text Parallel Corpus},
      booktitle = {The 51st Annual Meeting of the Association for Computational Linguistics - Short Papers},
      month     = {August},
      year      = {2013},
      address   = {Sofia, Bulgaria},
      publisher = {Association for Computational Linguistics},
      url       = {http://acl2013.org/site/short/2494.html}
    }
    

    Relevant Papers

  • Collective Generation of Natural Image Descriptions.
    Polina Kuznetsova, Vicente Ordonez, Alexander Berg, Tamara Berg and Yejin Choi.
    Association for Computational Linguistics (ACL), 2012 [bib][pdf]