As can be seen in SVN (http://tesseract-polish.googlecode.com/svn/trunk/src/training_images/boxtiff.pol/), currently all training data is based only on 300 DPI scans.

The cause for this is simple - when I've been working on training, I only had a 300 DPI scanner at my disposal.

This results in somewhat suboptimal recognition of texts printed/scanned in 600 DPI or more.

Now I have access to a 600 and 1200 DPI scanner, but have no time whatsoever to perform additional training using that.

If anyone would step up to do this work, it would be very appreciated.