June 2007
Volume 7, Issue 9
Free
Vision Sciences Society Annual Meeting Abstract  |   June 2007
Object and scene recognition in tiny images
Author Affiliations
  • Antonio Torralba
    Computer Science and Artificial Intelligence Laboratory, MIT
  • Rob Fergus
    Computer Science and Artificial Intelligence Laboratory, MIT
  • William T. Freeman
    Computer Science and Artificial Intelligence Laboratory, MIT
Journal of Vision June 2007, Vol.7, 193. doi:https://doi.org/10.1167/7.9.193
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Antonio Torralba, Rob Fergus, William T. Freeman; Object and scene recognition in tiny images. Journal of Vision 2007;7(9):193. https://doi.org/10.1167/7.9.193.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

The human visual system is remarkably tolerant to degradations in image resolution: in a scene recognition task, the performance of subjects is similar whether 32x32 color images or multi-mega pixel images are used. Even object recognition and segmentation is performed robustly by the visual system despite the object being unrecognizable in isolation. We present a set of studies to evaluate the minimal image resolution required to perform a number of recognition tasks (scene recognition, object detection and segmentation) and we show that images need 32x32 color pixels. Performances degrade fast below this resolution. The small size of each image carries two important benefits: (i) it permits computer vision tools to be easily applied and (ii) huge image databases may be easily collected. We present a database of 70,000,000 32x32 color images gathered from the Internet using image search engines. Each image is loosely labeled with one of the 70,399 non-abstract nouns in English, as listed in the Wordnet lexical database. Hence the image database represents a dense sampling of all semantic categories. Computer vision traditionally consider a few unrelated classes which are treated independently to one another. In contrast, we use our database in conjunction with a semantic hierarchy, obtained from Wordnet, to impose tree-structured dependencies between the 70,399 classes.

Torralba, A. Fergus, R. Freeman, W. T. (2007). Object and scene recognition in tiny images [Abstract]. Journal of Vision, 7(9):193, 193a, http://journalofvision.org/7/9/193/, doi:10.1167/7.9.193. [CrossRef]
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×