Hey!

Could someone give me a direction on this? I'm attempting to achieve
an over-time adaptation of a recognizer so that for example, I could
feed in pages (shopping bills in my case) and every time I notice
recognition errors, I'd correct them and let the system learn from it,
kind of incrementally. For one thing, I don't know how feasible my
approach would be and for another, I'm stuck!

So far I've got a page image and a transcription for it. I could also
make them into several line-text pairs if it's needed.

Where I'm stuck is in trying to generate cseg images, as I understood
they would be useful for training. I tried to make "align-
transcription.lua" to run but finally, after a few import tweaks I
noticed that function's
"recognizeLine(fst_builder,map,segmentation,image)" parameters didn't
have "map" anymore defined, but later in the script function
"ocr_result_to_charseg(cseg,map,ids,segmentation)" would've required
it. Could I fix it or should I try something else? For the "could"
part, I'd immediately need some help on how to do it, but if it was
something else, it might keep me quiet a while until stuck again. :)

>From there, I figured, I could then use "train-bpnet-lines.lua" to
make a new bpnet file and use it for recognizing from then on,
repeating the training whenever there's more errors.

Suggestions?

Thanks!
Caius

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"ocropus" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/ocropus?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to