Signs and Universal Text

Overview

Visually impaired individuals have achieved great autonomy using a combination of traditional aids and more recent advances suchs as GPS, reading devices for printed text, and other technologies. However, the desire or need to read street signs, store front banners, marquees, and other forms of text that are ubiquitous cannot yet be met without the aid of another person.


Our goal is to develop algorithms for robustly reading text in complex indoor and outdoor environments. We focus our efforts on three central issues:
  • Increased accuracy of detection and recognition.
  • Incorporating user input and goals.
  • Graceful failure to minimize harmful effects.

Our previous work is summarized on the prior VIDI Project web page.


Faculty


Graduate Students


Collaborators


References

  • Jerod J. Weinman and Erik Learned-Miller.
    Improving Recognition of Novel Input with Similarity
    IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2006.
  • Jerod J. Weinman, Allen Hanson, and Erik Learned-Miller.
    Fast Lexicon-Based Scene Text Recognition with Sparse Belief Propagation
    Intl. Conference on Document Analysis and Recognition (ICDAR), Sept. 2007.