Unfortunately, no. I wrote that I *want* to, but that's not quite
the same thing... Currently, I have to boot my second computer into
that dreaded win95 in order to do OCR under OmniPage Pro, which I
think is pretty good.
If anyone writes a twain driver that uses the netscane protocol,
I could try running OmniPage Pro under Wine (it *almost* worked
modulo the twain driver last time I tried, which was 5 or 6 months
ago) or under TWIN.
>Are you using xocr?
xocr is unforunately pretty unusable.
>I think it would be neat if there
>was good OCR support in SANE, but don't have experience with this
>myself (I tried xocr once, but it failed miserably on the Alpha, which
>I assumed was due to some 64 bit problems).
There's public-domain handwriting OCR software available from NIST.
It was paid for by (surprise, surprise) the US Census Bureau.
I've got a copy of it and am thinking of writing some software to
train it with common book fonts instead of handwriting; since it
is a neural net, that should be workable. I dont' know how big a
database it needs. I also don't know what kind of quality I could
get out of it. I'm thinking of using all the decent postscript
fonts I can get my hands on, and rendering them at different sizes
and skews and then using that to train with. The highest-level
parts of the NIST software assumes that you are reading a form,
but I think that rubber bands could define one-off "forms".
But I won't have any time to even consider playing with that until
"after the book is done"(TM). If anyone else is interested I can
look up the information. I'd guess that I'll be pretty unreliable
working on this project, since it's only one of many personal
interest projects I am considering working on.
"Ever wonder why the SAME PEOPLE make up ALL the conspiracy theories?"
-- Source code, list archive, and docs: http://www.azstarnet.com/~axplinux/sane/ To unsubscribe: mail -s unsubscribe firstname.lastname@example.org