Re: [orca-list] OCR

[chrys87 <chrys87 web de>]

Howdy,

maybe I m not at the current point. But my blind girlfrind use tesseract 
everyday and it seems to work pretty well for her. thats sayed.

> the performance is still difficult to use for "basic"
> users (I mean, the Windows users who don't have any technical knowledge
> or who use computer just for needs).

ohm OCR is never a short task. it also needs some time with windows.
tesseract could be trained as well for better results. i did this for 
our company and it works well for archive invoices.

> That's why I had a look at what provide proprietary world, waiting for
> having money enough to create a full OCR suite, free and based on
> Tesseract.

maybe i did not understand correctly. My girlfrind did try to use Abbyy 
under windows and its more and more less a11y under windows. If Lios is 
not a11y enought why not support them?
what are the needs of your users? OCR is not a simple task.

And we needed a quick solution, so that the GNU/Linux OS could be usable
by everyone now,

hm? why it is not useable by everyone now? My blind girlfrind did its 
whole study with linux. also OCR.
All the problems she had she talked to me to solve (and there years ago 
now). so what exact is the problem?

> * from an image file, right-click, choose the proper option

feel free to improve and integrate OCRpdf if they need a one click solution
https://github.com/chrys87/ocrpdf
but it seems that nobody cares about so i stopped the work.
its made to run it from the context menue. without many features right now.

> Waiting for a full "libre" solution, accessible for such people.

They are existing and may need to be improved in some situations but why 
to reexplore anything.

maybe i just misunderstand but tesseract is no rocket science to train 
and the project managers for sure looking for this. (because they are 
really active).

tesseract leads to good results here in a company with 60000 employees 
so what is the problem for private people?

could you explain the problem to me? I dont say its wrong, but based on 
my informations i just dont understand.
cheers chrys

Am 27.04.2016 um 19:58 schrieb MENGUAL Jean-Philippe:
> Hi,
>
> After test of various OCR, I feel that Tesseract, the most advanced OCR
> engine on Linux, hasn't noawadays all ways to be as performant as
> commercial utilities. Even if it's wrapped in some tools like Lios
> or gimagereader, the performance is still difficult to use for "basic"
> users (I mean, the Windows users who don't have any technical knowledge
> or who use computer just for needs).
>
> That's why I had a look at what provide proprietary world, waiting for
> having money enough to create a full OCR suite, free and based on
> Tesseract. Create or improve, as Lios and gimagereader are
> excellent points of beginning, but some things are hard to understand
> for our users in GUI (after tests).
>
> And we needed a quick solution, so that the GNU/Linux OS could be usable
> by everyone now, including OCR matter, so that they buy service and
> finance our devs projects for free software. But I wonder now if some
> usual GNU/Linux users here could be interested by such a product. What
> we reach now is a suite for 200E, including:
> - Abbyy FindReader 11, unlimited in number of pages thanks to an
> agreement between Abbyy and Hypra based on the fact we do a free program
> and designed for blind people with specifific needs in OCR,
> - A package to run it on MATE. 2 ways:
> * from an image file, right-click, choose the proper option
> * from a scanner: we give a command to create a binding (as ours in
> linked against Compiz).
>
> I precise that the utility could also use Tesseract if FindReader is
> missing, but in such case, it will be free.
>
> Would some users interested by such solution? I "like" it as it
> introduces OCR on GNU/Linux and enable some unusual users to come.
> Waiting for a full "libre" solution, accessible for such people.
>
> Regards,