Paperwork (https://github.com/jflesch/paperwork) is a great tool too. I'd take a look at lector. Been trying to get a sane (heh) ocr setup for a while.
Does paperwork give you a nice interface for training? The models that tesseract ship with are fairly terrible on everything I've tried them with, but add in 5 pages of training data and it becomes great.
4 comments
[ 2.8 ms ] story [ 18.9 ms ] thread