jocr-devels

[Jocr-devels] gocr for license plate recognition

[Stagecoach W. <bea...@st...>]

I am looking for experienced gocr/ jocr user/ developers to participate in a
license plate recognition development project.

A specification framework is available but needs technical input.

Please let me know if you are interested in such a project.

With kind regards
Beat Naef

Re: [Jocr-devels] Arabic numbers OCR

[grok <gr...@re...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


> I need an application that can look at a file and ignore
> everything but the numbers. The numbers are in Arabic. So
> pretty much convert scanned numbers to text numbers in Arabic.
> Can GOCR do this?
> 
> If not is anybody interested to get this going? I would be
> more than happy to help.

I'm not involved in development here whatsoever. However I just
wanted to opine that IMO, this is likely exactly the kind of
project JOCR would excel in. Just that you would need to be
aiming at a whole different set of characters than roman script,
eh?

Best of luck here. Arabic numerals can't be so hard. People want
to read licence plates off fotos of cars and such, for instance.
;>


- -- grok.







- -- 
*** FULL-SPECTRUM DOMINANCE! ***************************************
*          Boycott the Bourgeois Economy: BUY PROGRESSIVE          *
****** Critical endorsement only ***** Gift-giving Year-round ******
* http://www.oceanbooks.com.au                         Ocean Books *
* http://www.heritagebooksmusic.com HERITAGE BOOKS & MUSIC (Afrikan)
* http://www.zatoun.com   Fair Trade Extra Virgin Olive Oil & Soap *
* http://www.thebasketcasegifts.ca Six Nations Indian Reserve ON CA*
********************************************************************
GPG fingerprint = 2E7F 2D69 4B0B C8D5 07E3  09C3 5E8D C4B4 461B B771
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)

iD8DBQFJikFKXo3EtEYbt3ERAiriAKDyIQqTPgCKkdYd22aopkPi4NpfDwCgma/I
HrOuLw67qN5dQCLzEAtKFb8=
=0ifa
-----END PGP SIGNATURE-----

[Jocr-devels] Arabic numbers OCR

[ali r. <ali...@gm...>]

Hi there,
I am new to the list so apologies in advance if I am doing things wrong.

I need an application that can look at a file and ignore everything but the
numbers. The numbers are in Arabic. So pretty much convert scanned numbers
to text numbers in Arabic. Can GOCR do this?

If not is anybody interested to get this going? I would be more than happy
to help.

Thank you very much in advance,

Ali


-- 
CONFIDENTIALITY: This email (including any attachments) may contain
confidential, proprietary and privileged information, and unauthorised
disclosure or use is prohibited. If you received this email in error, please
notify the sender and delete this email from your system.

Re: [Jocr-devels] OCRing 7-segment digits

[Nikolaus R. <Nik...@ra...>]

Carl Karsten <ca...@pe...> writes:
> Nikolaus Rath wrote:
>> Hello,
>> 
>> I am trying to OCR 7-segment digits like this:
>> http://www.rath.org/ocr.gif (actually they're not quite 7 segment
>> digits because they also include a decimal point). I expect that in
>> theory it should be quite easy to recognize them correctly, but jocr
>> with its default settings fails dramatically (probably because it
>> tries to match a "normal" font).
>> 
>> Is there an easy way to make jocr recognize these images?
>
> I asked this exact thing back in June or July of 08. I got some
> pretty good sounding advice, but I was done needing it so I never
> got to try it out. search the logs, let me know if you can't find
> anything about it.

I think I found the relevant thread, but there was only this
suggestion:

,----
| read pictures exiv for datestamp
| crop it to digits only
| (remove dot)
| black and white
| gorc
| -> database
`----

But this is already what I do - only that in my case gocr does not
recognize the digits and that I would like to keep the decimal point.
Unfortunately your sample picture wasn't online anymore, so I couldn't
compare it to my data.


Best,


   -Nikolaus

-- 
 »Time flies like an arrow, fruit flies like a Banana.«

  PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6  02CF A9AD B7F8 AE4E 425C

Re: [Jocr-devels] OCRing 7-segment digits

[Carl K. <ca...@pe...>]

Nikolaus Rath wrote:
> Hello,
> 
> I am trying to OCR 7-segment digits like this:
> http://www.rath.org/ocr.gif (actually they're not quite 7 segment
> digits because they also include a decimal point). I expect that in
> theory it should be quite easy to recognize them correctly, but jocr
> with its default settings fails dramatically (probably because it
> tries to match a "normal" font).
> 
> Is there an easy way to make jocr recognize these images?

I asked this exact thing back in June or July of 08.  I got some pretty good 
sounding advice, but I was done needing it so I never got to try it out.  search 
the logs, let me know if you can't find anything about it.

Carl K

[Jocr-devels] OCRing 7-segment digits

[Nikolaus R. <Nik...@ra...>]

Hello,

I am trying to OCR 7-segment digits like this:
http://www.rath.org/ocr.gif (actually they're not quite 7 segment
digits because they also include a decimal point). I expect that in
theory it should be quite easy to recognize them correctly, but jocr
with its default settings fails dramatically (probably because it
tries to match a "normal" font).

Is there an easy way to make jocr recognize these images?


Best,


   -Nikolaus

-- 
 »Time flies like an arrow, fruit flies like a Banana.«

  PGP fingerprint: 5B93 61F8 4EA2 E279 ABF6  02CF A9AD B7F8 AE4E 425C

[Jocr-devels] gocr and database

[flo <fl...@ma...>]

Hello,
I'm using gocr 0.41-1 on debian to convert a png file into txt file.
My source image is black text on white background.
I attach the source image to this mail.

First I have run the simple command : gocr input.png -o output.txt.
I notice a problem with some characters. Then I run the command line : 
gocr -a 100 -p ~/tmp/db/dbACM/ -m 130 input.png -o output.txt

I give an answer for all unrecognized characters and tell gocr to store
the information in the database directory.

Then I run the command : 
gocr -a 100 -p ~/tmp/db/dbACM/ -m 2 input.png -o output.txt

I was hoping that gocr will load the information from the database
directory but it recognizes letters as if their is nothing in the
database, like when I run the command line : 
gocr input.png -o output.txt

So can you please tell me if I made a mistake using gocr.
Do you have any solution to work with the database ?

Thank you for your help,

F.J.

[Jocr-devels] Re : Re : problem on l and '

[Pluch' <plu...@ya...>]

That was a wonderful idea for this wonderful program.
Adding white borders didn't solve the problem, but when drawing a smaller apostropheit worked :
l : 3*28 px
apostrophe : 2*7 px 

2*5 px didn't work , maybe a proporionally problem as you said.
Thanks again

Pluch'






----- Message d'origine ----
> De : AJ <abo...@pa...>
> À : Pluch' <plu...@ya...>; joc...@li...
> Envoyé le : Jeudi, 8 Janvier 2009, 19h17mn 38s
> Objet : Re : [Jocr-devels] problem on l and '
> 
> The image sizes I always used were a consistent hight.  So, if in
> normal use of the font, the apostrophe has lots of whitespace under
> it, then that's how I trained it.  Also, I always trained with some
> whitespace on both sides of the character. When it trains the
> apostrophe, does it try to autosize the its own box?  Maybe it ends
> up cutting off all the whitespace at the bottom and then training the
> shape, which would probably be exactly like the "l" proportionally.
> Is there a way to force it to use a certain box size so it does look
> at the whitespace?
> 
> > Hello ans thanks for your response,
> >
> > Yes, I've done that too (the images are : apostrophe 2*10 px and the l is 3*28 
> px) and if  the l is before the apostrophe in the database, both characters are 
> l (if the apostrophe is before l, both are apostrophes).
> >
> > I tried with a 2*28 px image for the apostrophe (10 px height black and 18 px 
> white) but this didn't solve the problem ...
> >
> > Any other clue is welcome !
> >
> > Thks
> >
> >
> >
> >
> >
> >
> > ----- Message d'origine ----
> > > De : AJ 
> > > À : Pluch' ; joc...@li...
> > > Envoyé le : Jeudi, 8 Janvier 2009, 18h38mn 43s
> > > Objet : Re: [Jocr-devels] problem on l and '
> > >
> > > Havent used GOCR in a while, but have you tried training the the
> > > apostrophe and the letter "l" separately?  I vaguely remember GOCR
> > > having trouble training on multiple letter sequences, so I always
> > > trained one character at a time.
> > >
> > > > Hello,
> > > >
> > > > I hope I am on the right place for the question bellow.
> > > >
> > > > With gocr 0.41, I try to build my own database, but there is a problem 
> with
> > > the apostrophe and the letter l.. With the text "S'il" gocr ask me for the
> > > apostrophe but not for the l : for gocr, the l is like an apostrophe :
> > > >
> > > > Here is what I get with the command line :
> > > >
> > > > gocr -p /home/jeanclaudetergal/txt2/db/ -f UTF8 -v 32  -m 442 -i img.bmp
> > > >
> > > > #DEBUG: ocr_db (285,7)
> > > > # show box + environment
> > > > # show box     x=  285    7 d=   2  10 r= 0 0
> > > > # show pattern x=  260    5 d=  52  37 t= 1 1
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
> > > > ,,,,,OOOOOOOOOO,,,,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,, -
> > > > ,,,OOOOOOOOOOOOOO,,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,,<
> > > > ,,OOOOOOOOOOOOOOOO,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,OOOOOO,,,,,,OOOOOO,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,OOOO,,,,,,,,,,OOOO,,,,,,##,,,,,,,,,,,,,,OOO,,,,,,,,
> > > > OOOO,,,,,,,,,,,,OOOO,,,,,##,,,,,,,,,,,,,,OOO,,,,,,,,
> > > > OOOO,,,,,,,,,,,,OOOO,,,,,##,,,,,,,,,,,,,,OOO,,,,,,,,
> > > > OOOO,,,,,,,,,,,,,OO,,,,,,##,,,,,,,,,,,,,,OOO,,,,,,,,
> > > > OOOO,,,,,,,,,,,,,,,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,,<
> > > > OOOOO,,,,,,,,,,,,,,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,, -
> > > > OOOOOO,,,,,,,,,,,,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,OOOOOOOO,,,,,,,,,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,OOOOOOOOOOOO,,,,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,,OOOOOOOOOOOOOO,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,,,,,OOOOOOOOOOOO,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,,,,,,,,,OOOOOOOOO,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,,,,,,,,,,,,,OOOOOO,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,OOOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > OO,,,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > OOO,,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > OOO,,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > OOOO,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > OOOOO,,,,,,,,,,,OOOO,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,OOOOO,,,,,,,,OOOOOO,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,OOOOOOOOOOOOOOOOO,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,,OOOOOOOOOOOOOO,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,
> > > > ,,,,,OOOOOOOOOO,,,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,<
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
> > > > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,<
> > > > The above pattern was not recognized.
> > > > Enter UTF8 char or string for above pattern. Leave empty if unsure.
> > > > (I hope this is readable)
> > > >
> > > > You can see the original image here : http://scriptus.free.fr/img/img.bmp..
> > > > What do you think I can do with this problem ? Should I change some 
> parameters
> > > or there is no issue with this fonts ?
> > > >
> > > > Thanks in advance and Happy New Year!
> > > >
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > 
> ------------------------------------------------------------------------------
> > > > Check out the new SourceForge.net Marketplace.
> > > > It is the best place to buy or sell services for
> > > > just about anything Open Source.
> > > > http://p.sf.net/sfu/Xq1LFB
> > > > _______________________________________________
> > > > Jocr-devels mailing list
> > > > Joc...@li...
> > > > https://lists.sourceforge.net/lists/listinfo/jocr-devels
> >
> >
> >
> >
> >



      

[Jocr-devels] problem on l and '

[Pluch' <plu...@ya...>]

Hello,

I hope I am on the right place for the question bellow.

With gocr 0.41, I try to build my own database, but there is a problem with the apostrophe and the letter l. With the text "S'il" gocr ask me for the apostrophe but not for the l : for gocr, the l is like an apostrophe :

Here is what I get with the command line :

gocr -p /home/jeanclaudetergal/txt2/db/ -f UTF8 -v 32  -m 442 -i img.bmp

#DEBUG: ocr_db (285,7) 
# show box + environment
# show box     x=  285    7 d=   2  10 r= 0 0
# show pattern x=  260    5 d=  52  37 t= 1 1
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,  
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,  
,,,,,OOOOOOOOOO,,,,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,, -
,,,OOOOOOOOOOOOOO,,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,,< 
,,OOOOOOOOOOOOOOOO,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,,  
,OOOOOO,,,,,,OOOOOO,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,,  
,OOOO,,,,,,,,,,OOOO,,,,,,##,,,,,,,,,,,,,,OOO,,,,,,,,  
OOOO,,,,,,,,,,,,OOOO,,,,,##,,,,,,,,,,,,,,OOO,,,,,,,,  
OOOO,,,,,,,,,,,,OOOO,,,,,##,,,,,,,,,,,,,,OOO,,,,,,,,  
OOOO,,,,,,,,,,,,,OO,,,,,,##,,,,,,,,,,,,,,OOO,,,,,,,,  
OOOO,,,,,,,,,,,,,,,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,,< 
OOOOO,,,,,,,,,,,,,,,,,,,,##,,,,,OOO,,,,,,OOO,,,,,,,, -
OOOOOO,,,,,,,,,,,,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,OOOOOOOO,,,,,,,,,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,OOOOOOOOOOOO,,,,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,,OOOOOOOOOOOOOO,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,,,,,OOOOOOOOOOOO,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,,,,,,,,,OOOOOOOOO,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,,,,,,,,,,,,,OOOOOO,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,,,,,,,,,,,,,,,OOOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,,,,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
OO,,,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
OOO,,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
OOO,,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
OOOO,,,,,,,,,,,,,OOOO,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
OOOOO,,,,,,,,,,,OOOO,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,OOOOO,,,,,,,,OOOOOO,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,OOOOOOOOOOOOOOOOO,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,,OOOOOOOOOOOOOO,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,  
,,,,,OOOOOOOOOO,,,,,,,,,,,,,,,,,OOO,,,,,,OOO,,,,,,,,< 
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,  
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,  
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,  
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,  
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,  
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,  
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,< 
The above pattern was not recognized.
Enter UTF8 char or string for above pattern. Leave empty if unsure.
(I hope this is readable)

You can see the original image here : http://scriptus.free.fr/img/img.bmp.
What do you think I can do with this problem ? Should I change some parameters or there is no issue with this fonts ?

Thanks in advance and Happy New Year!



      

Re: [Jocr-devels] Real-world images

[Joerg S. <Joe...@UR...>]

There is a plan for such things on my ToDo list but not more.
I hope I have more time for coding next year.

Joerg

On Mon, 22 Dec 2008, Calder, Erick wrote:

> could you guys give me an assessment of how useful this library would be
> for doing recognition on real-time images? I'm interested in being able
> to recognize certain types of vehicles from images taken from a freeway.
>
>
>
> alternatively, could anyone suggest a library better suited for such a
> task?
>
>
>
> thx! - erick
>
>
>
> Attention: This email is intended for the sole use of the intended recipient(s) and may contain confidential or privileged information. No one is authorized to copy or re-use this email or any information contained in it. If you are not the intended recipient, we request that you please notify us by reply email and destroy all copies of the message and any attachments. Thank you for your cooperation.
>
>

-- 
-------------------------------------------------------------------------
-                                                             \V/       -
- EMAIL: joe...@ur...              (o o)      -
----------------------------------------------------------oOo-(_)-oOo----
- http://www-e.uni-magdeburg.de/jschulen/
- PGP 1024D/53BDFBE3, 3816 B803 D578 F5AD 12FD  FE06 5D33 0C49 53BD FBE3
-------------------------------------------------------------------------

[Jocr-devels] Real-world images

[Calder, E. <Eri...@tc...>]

could you guys give me an assessment of how useful this library would be
for doing recognition on real-time images? I'm interested in being able
to recognize certain types of vehicles from images taken from a freeway.

 

alternatively, could anyone suggest a library better suited for such a
task?

 

thx! - erick



Attention: This email is intended for the sole use of the intended recipient(s) and may contain confidential or privileged information. No one is authorized to copy or re-use this email or any information contained in it. If you are not the intended recipient, we request that you please notify us by reply email and destroy all copies of the message and any attachments. Thank you for your cooperation.

[Jocr-devels] libgocr

[Marcin W. <mwy...@gm...>]

hello i have tiny question. I'm writing a master thesis about car license
plates. I have to use lib of gocr in my project could you tell me how to use
jocr to read a cut fragment from photo on which we have only a car license
plate? i would be grateful to show me an example :) thx a lot :)

-- 
Pozdrawiam Marcin Wyczechowski

[Jocr-devels] Barcode support in gocr?

[Martin T. <ma...@if...>]

Hello,

I am trying to use gocr to read barcodes in scanned images. Type 39
works fine, but it seems gocr does not support Codabar? Unfortunately,
I need this. I am willing to try to make a patch, but I would rather
not duplicate work if anyone else have already looked into this?

I would also like a switch to tell gocr to only look for barcodes;
however, I've found a tip to use -m 312, and in addition I can crop
the image (the barcode I'm interested in is in a specified place).
This seems to speed things up enough for the time being.


Thanks,

Martin
-- 
"An ideal world is left as an exercise to the reader."
                                                 -Paul Graham, On Lisp

[Jocr-devels] Speicfying Box/Pattern Possible?

[Joe H. <de...@in...>]

Is it possible to specify the box/pattern ?   I'm finding that its often
choosing boxes/patterns that arent at all accurate.   Like for example in
this one image of the letter 'M', its just choosing a smasll part of the
letter M, and I want some way to manually specify a larger pattern/box
that covers the entire letter M.   and then tell it thats the letter M.


I'm working on training GOCR, and currently I've gone as far as to split
up images into tons of small images containing a single letter in each
image, and then the plan was to just train it what each letter was,
however in many image files its not selecting the entire letter as the
pattern, only a tiny fragment of the letter.

If I worded this poorly I'm sorry.

If you cant manually specify regions for a letter in GOCR, is there
another program that can?


-- 
HELLO! My name is Inigo Montoya, you killed my father, prepair to die!

[Jocr-devels] black background issue

[Carl K. <ca...@pe...>]

When I do a screen shot of a VM in 80x25 white (or green) on black text mode,
then gocr it, it is completely lost, unless I flip the colors:

carl@dell29:~/a$ import vnc1.png
carl@dell29:~/a$ gocr vnc1.png
____>_0___00__ _0 ,___0____,___


_______ ,,___E3_____E3__E3_____,___E3_0__ _E3E3__0____ __0_____
______ ______ __0__ __________0__ _____ _________0
i___0 ____,,, ____0 _ 0_
____>_0___00___0

carl@dell29:~/a$ convert -negate vnc1.png vnc1a.png

carl@dell29:~/a$ gocr vnc1a.png
carI_ibexx64:~$ ./MXtun.sh

Malloc:  ../bash/dispose_cMd.c:241: assertion botched
free: called with unallocated blocX arguMent
Aborting...Aborted
carI_ibexx64:~$ _

Also, what I can do to help train gocr to recognize this text 100%?

Carl K

[Jocr-devels] libPgm2asc.a in gocr-0.46

[Igor F. [Contr] <ig...@he...>]

Dear GOCR developers,

First of all I would like to say thank you for the great job you are
doing! I have tried/tested OCRAD, Tesseract and a few other OCR programs
and while all of them have their strengths and weaknesses to my
knowledge GOCR is the only project that has a clean and simple API to
use from  another GPL project.
Which is why it was to my great dismay I have found that in the latest
version - 0.46 the generation of libPgm2asc.a has been discontinued "for
simplicity".  I am a developer of open source optical chemical structure
recognition software - OSRA (http://osra.sourceforge.net). I rely on
GOCR library for character recognition as part of my program. Removing
libPgm2asc is an inconvenience at the least and a major problem at the
worst. Calling the reason "for simplicity" is just adding insult to
injury. For right now I'm recommending the users of my software to avoid
gocr-0.46 but this situation is untenable in the long run. I am urging
to change this and restore the creation of libPgm2asc library as this is
the  piece of GOCR that is most important to my project and this is what
distinguishes GOCR from all other OCR software.

Best regards,
Igor Filippov

Re: [Jocr-devels] possible idea to increase chance of 0's instead of O's

[Joerg S. <Joe...@UR...>]

The context correction was to weak. I fixed at in 
www.ovgu.de/jschulen/jocr.tgz (its not official released yet).
You can limit to numbers and decimal point using -C "0-9.," option.
If you think gocr does stupid things, please email a small sample to me
to give me the chance to fix the problem.

Regrads,
  Joerg.


On Sun, 19 Oct 2008, Sam P wrote:

> A lot of the times gocr misinterprets 5's as S's, 1's as l's, and 0's as
> o's.
> If surrounding characters were looked at then it would be more probable that
> a numbers surround other numbers.
>
> I use gocr to read in coordinate measurement reports and am quite impressed
> with it.  I tried 2 other open source ocr programs and gocr was by far the
> easiest to use with the best results.
>
> To overcome the numbers I limited the results to numerical values only,
> although it sometimes misses a decimal point.
>
> Any hints / guides on how to tweak the values?
>
> Thank you,
> Sam Price

[Jocr-devels] possible idea to increase chance of 0's instead of O's

[Sam P <fl...@gm...>]

A lot of the times gocr misinterprets 5's as S's, 1's as l's, and 0's as
o's.
If surrounding characters were looked at then it would be more probable that
a numbers surround other numbers.

I use gocr to read in coordinate measurement reports and am quite impressed
with it.  I tried 2 other open source ocr programs and gocr was by far the
easiest to use with the best results.

To overcome the numbers I limited the results to numerical values only,
although it sometimes misses a decimal point.

Any hints / guides on how to tweak the values?

Thank you,
Sam Price

Re: [Jocr-devels] recognizing 7 segment numbers

[Carl K. <ca...@pe...>]

awesome!  thanks.

Know of any python bindings for gocr?  i could use PIL for the image stuff, and 
the -> database step would be easier.

Carl K

Christian Stern wrote:
> I tested it out.
> 
> gorc was able to read the digits. i cropped the picture. turned in
> black and white and removed the dot. if the cam is on a fixed position
> it is possible to put all these in a script to get your volt-display
> into a file or database.
> 
> script:
> 
> read pictures exiv for datestamp
> crop it to digits only
> (remove dot)
> black and white
> gorc
> -> database
> 
> If there is no contant brightness and a window you will need some
> settings fit to the daylight outside.
> 
> possible!
> 
> Christian Stern
> 
> 2008/7/23 Carl Karsten <ca...@pe...>:
>> I have a webcam pointed at a digital volt meeter - it saves a jpg every minute.
>>  I want to graph the numbers over time (time stamp of the file) - here is a sample:
>>
>> http://dev.personnelware.com/carl/temp/Jul23/a/motion/01-20080723093400-snapshot.jpg
>>
>> I am guessing this is outside what gocr is meant for, but figured I would ask,
>> and ask for suggestions too.
>>
>> one thought is I can probably check for which segments are on, and then it is a
>> simple lookup: assign segments to bits, 0 is the 6 outer segments on, so
>> something like 1111110 = 0x7e = 126.  1 is just the 2 on the right: 0110 0000 -
>> 0x60 = 96.
>>
>> 126, 0
>> 96, 1
>> ...
>> 127, 8
>>
>> The tricky part is identifying where the segments are.  for the most part
>> everything is stationary, but as I re-configure my tests I move the cam/meeter
>> around, so each run will have the segments in different parts of the image.
>> If gocr could somehow help with this, yay.
>>
>> Carl K
>>
>>
>>
>>
>>
>> -------------------------------------------------------------------------
>> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
>> Build the coolest Linux based applications with Moblin SDK & win great prizes
>> Grand prize is a trip for two to an Open Source event anywhere in the world
>> http://moblin-contest.org/redirect.php?banner_id=100&url=/
>> _______________________________________________
>> Jocr-devels mailing list
>> Joc...@li...
>> https://lists.sourceforge.net/lists/listinfo/jocr-devels
>>
> 
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
> Build the coolest Linux based applications with Moblin SDK & win great prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> Jocr-devels mailing list
> Joc...@li...
> https://lists.sourceforge.net/lists/listinfo/jocr-devels
> 
> 

Re: [Jocr-devels] recognizing 7 segment numbers

[Christian S. <der...@go...>]

I tested it out.

gorc was able to read the digits. i cropped the picture. turned in
black and white and removed the dot. if the cam is on a fixed position
it is possible to put all these in a script to get your volt-display
into a file or database.

script:

read pictures exiv for datestamp
crop it to digits only
(remove dot)
black and white
gorc
-> database

If there is no contant brightness and a window you will need some
settings fit to the daylight outside.

possible!

Christian Stern

2008/7/23 Carl Karsten <ca...@pe...>:
> I have a webcam pointed at a digital volt meeter - it saves a jpg every minute.
>  I want to graph the numbers over time (time stamp of the file) - here is a sample:
>
> http://dev.personnelware.com/carl/temp/Jul23/a/motion/01-20080723093400-snapshot.jpg
>
> I am guessing this is outside what gocr is meant for, but figured I would ask,
> and ask for suggestions too.
>
> one thought is I can probably check for which segments are on, and then it is a
> simple lookup: assign segments to bits, 0 is the 6 outer segments on, so
> something like 1111110 = 0x7e = 126.  1 is just the 2 on the right: 0110 0000 -
> 0x60 = 96.
>
> 126, 0
> 96, 1
> ...
> 127, 8
>
> The tricky part is identifying where the segments are.  for the most part
> everything is stationary, but as I re-configure my tests I move the cam/meeter
> around, so each run will have the segments in different parts of the image.
> If gocr could somehow help with this, yay.
>
> Carl K
>
>
>
>
>
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
> Build the coolest Linux based applications with Moblin SDK & win great prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> Jocr-devels mailing list
> Joc...@li...
> https://lists.sourceforge.net/lists/listinfo/jocr-devels
>

[Jocr-devels] recognizing 7 segment numbers

[Carl K. <ca...@pe...>]

I have a webcam pointed at a digital volt meeter - it saves a jpg every minute. 
  I want to graph the numbers over time (time stamp of the file) - here is a sample:

http://dev.personnelware.com/carl/temp/Jul23/a/motion/01-20080723093400-snapshot.jpg

I am guessing this is outside what gocr is meant for, but figured I would ask, 
and ask for suggestions too.

one thought is I can probably check for which segments are on, and then it is a 
simple lookup: assign segments to bits, 0 is the 6 outer segments on, so 
something like 1111110 = 0x7e = 126.  1 is just the 2 on the right: 0110 0000 - 
0x60 = 96.

126, 0
96, 1
...
127, 8

The tricky part is identifying where the segments are.  for the most part 
everything is stationary, but as I re-configure my tests I move the cam/meeter 
around, so each run will have the segments in different parts of the image. 
If gocr could somehow help with this, yay.

Carl K

Re: [Jocr-devels] Divide 2 digits.

[Emanoil K. <del...@ya...>]

sorry for mixing into this but I don't agree its good project, because:

1) it supports only latin chars
2) it does not use vectors
3) it is not able to learn - kind of stochastics missing

NO one gave an answer at least to the question how it can be done to support cyrillic chars etc.

If you try t odo something useful better do not relay on gocr and take a commercial tool - that's what I did.

I'm waiting to see what the google ocr will be able to do

regards


--- On Tue, 7/22/08, Christian Stern <der...@go...> wrote:

> From: Christian Stern <der...@go...>
> Subject: [Jocr-devels] Divide 2 digits.
> To: joc...@li...
> Date: Tuesday, July 22, 2008, 4:18 PM
> First of all, gocr is a great tool. Fast to learn and
> finally to use it.
> 
> But some Questions left.
> 
> running gocr with:
> 
> > gocr -p . -C 0-9 -s -1 -m 178 lesemich.pnm
> 
> After some minutes, the Database is able to read almost
> every scanned
> paged correctly. Some times two digits are so close
> together and the
> program is asking because digits are not recognized.
> 
> Example:
> 
> 53 is been not recognized.
> 53 is marked
> 
> What should i type in.
> 
> 53
> or
> 5 3
> or something else?
> 
> Same Questions to more then 2 digits (happens every Page
> full of
> digits ~ 5 times)
> 
> - Possible to import fonts, font-set.
> - Good idea to merge 2 databases. Both been working for
> some days, but
> on different pages with differents and content.
> 
> Somtimes the enginge got problems at 0, 3, 8. Does it make
> sense to do
> runs like.
> > gocr  -C 0 ...
> 
> to only "learn" these character?
> 
> Christian Stern
> 
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move
> Developer's challenge
> Build the coolest Linux based applications with Moblin SDK
> & win great prizes
> Grand prize is a trip for two to an Open Source event
> anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> Jocr-devels mailing list
> Joc...@li...
> https://lists.sourceforge.net/lists/listinfo/jocr-devels


      

[Jocr-devels] Divide 2 digits.

[Christian S. <der...@go...>]

First of all, gocr is a great tool. Fast to learn and finally to use it.

But some Questions left.

running gocr with:

> gocr -p . -C 0-9 -s -1 -m 178 lesemich.pnm

After some minutes, the Database is able to read almost every scanned
paged correctly. Some times two digits are so close together and the
program is asking because digits are not recognized.

Example:

53 is been not recognized.
53 is marked

What should i type in.

53
or
5 3
or something else?

Same Questions to more then 2 digits (happens every Page full of
digits ~ 5 times)

- Possible to import fonts, font-set.
- Good idea to merge 2 databases. Both been working for some days, but
on different pages with differents and content.

Somtimes the enginge got problems at 0, 3, 8. Does it make sense to do
runs like.
> gocr  -C 0 ...

to only "learn" these character?

Christian Stern

[Jocr-devels] few questions

[Emanoil K. <del...@ya...>]

Guys, I'm glad to see that you are alive and still working on the project.

When can we have finally support for cyrillic?

You were also talking of moving to vector processing and not pixel - did you do this? Are you planning and when?

thanks and regars



      

[Jocr-devels] patch for the gocr api

[Patrice D. <per...@fr...>]

Hello,

Here is a patch against the tarball which seemed to me to be the most
recent one to clean up a bit the API. It allows to have 
gocr.h list.h pnm.h unicode.h pgm2asc.h
for the API, with as few as conditionals as possible.

There is still one conditional, in fact, HAVE_GETTIMEOFDAY, I don't
know how to get rid of it, it would be nice to avoid it in the API.
Could you do it? 

I put the debug define in a file I named debug.h I also attach.

I also added some includes when there were warnings.

PS: I am ont subscribed, so please put me in CC.

--
Pat