Jason (jcreed) wrote,

Does anybody know about any prior work using off-the-shelf machine learning algorithms in typography to learn how sidebearings and kerning are a function of raw (construed as translation-independent) glyph data? I did a few google searches and turned up nothing. It seems like really low-hanging fruit, since every existing font could be used as training data. Plausible features I can think of off the top of my head are greatest extent left and right for various particular y-values, approximations of inter-glyph counter area for kerning, known kerning classes - or, come to think of it, maybe some sort of clustering approach could be used to justify (or refute) common rules of thumb about kerning classes.
Tags: fonts, math
  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded 

  • 1 comment