Ah, yes, one of the big features we are waiting on from research. We had a look at a couple Open Source OCR systems in the past (I'm sure they are probably better today) but found that for our usage (i.e. real world images) and not scanned paper images they would not work reliably. The main reason is the translation of a letter can undergo in the real world. What many of the OCR packages require is a frontal view of the letter. Think of looking at a road sign. Would you like the requirement of being directly in front of the sign in order to read it? Shouldn't standing an angle of 45 deg off center to the sign still be legible? We think so and most robots or real world applications require reading text from non-frontal (i.e. affine/perspective transfomed) text. But that's a much tricker task to accomplish than just straight template matching like what Tesserac does.
And we didn't get into Z plane rotation ...
So as there are Open Source packages out there that will do OCR for standard cases we have many ways that RR can interact with those packages that don't require a direct module (think just saving an image to disk and using the program execute module). Otherwise, we'd be writing a LOT of modules for a LOT of Open Source packages.
Or if there is a package you really like, we can help you to get started in writing a module to interface to it.
Otherwise, we have a lot done already on our own OCR module but its just not ready for primetime use ... seems that the W's and V's are causing alignment problems ... so if you don't need to read those letters we're all set! (just kidding!).
Thanks for the note, this helps us to keep pressure on getting this module out! Hopefully sometime soon ...