Windows Tool to edit boxfiles

[unknowner]

Hi guys,
if your going to train an new language this tool might help you a bit.
It shows the generated boxefiles over the picture.

you need:
- dot net 2.0
- picture as uncompressed tif
- boxfile (tesseract.exe train.tif train nobatch makebox) in the same
dir as the picture
- both files need to have the same name (train.tif train.txt)
- start bbTesseract
> open txt
> you can edit, delete, add new boxes
-Save boxfile

Give it a try and let me know it somethings going wrong. Don't await
to much as its V0.000000 ... 02 :)

download it here >> http://www.struthers.de/david/bbTesseract.exe

cu

[unknowner]

You can also download it from the groups files
http://groups.google.com/group/tesseract-ocr/web/bbTesseract.exe.gz

[unknowner]

Small update V0.00000...3 forgot to save as Unicode

[withbl...@gmail.com]

Hi,
Before receipt of this email, just downloaded. Whether download again? Whether it works for Kannada (one of the Indian lang)?
How to download dot net2.0  it would have nice  this also included in exe package
I am using xp.  If you want sample image of Kannada, I shall email you on hearing from you.
-74 yrs old

[unknowner]

- just download again and replace exe.
- you can download .net form microsoft ->
http://www.microsoft.com/downloads/details.aspx?displaylang=en&FamilyID=0856eacb-4362-4b0d-8edd-aab15c5e04f5
- send me your image, so i can do a some testing

On 31 Jul., 17:09, withblessi...@gmail.com wrote:
> Hi,
> Before receipt of this email, just downloaded. Whether download again?
> Whether it works for Kannada (one of the Indian lang)?
> How to download dot net2.0 it would have nice this also included in exe
> package
> I am using xp. If you want sample image of Kannada, I shall email you on
> hearing from you.
> -74 yrs old
>

[unknowner]

Received you mail, and it seems the error message is because you
haven't installed .net framework from microft (see link above).

On 31 Jul., 17:16, unknowner <struthers.da...@googlemail.com> wrote:
> - just download again and replace exe.

> - you can download .net form microsoft ->http://www.microsoft.com/downloads/details.aspx?displaylang=en&Family...

[Keith Beaumont]

Is this  a windows .exe? Don't think so. Can you compile one? If not can you release the source? Can anybody else then pick this up & do the necessary?

KB

 

On 7/31/07, unknowner <struthe...@googlemail.com> wrote:

[Keith Beaumont]

PS. Error msg is "The application failed to initialise properly (0xc0000135)"

[Keith Beaumont]

PPS. This couldn't be because I don't have dot net 2.0?

[unknowner]

hi keith,
it is a windows exe and it requires dot net framework ( ->thats your
error 0xc0000135.) Sorry for that.

This tool was made with VisualBasic 2005 Express, if you still want
the Source will upload it.

On 31 Jul., 20:11, "Keith Beaumont" <beaumon...@gmail.com> wrote:
> PPS. This couldn't be because I don't have dot net 2.0?
>

> On 7/31/07, Keith Beaumont <beaumon...@gmail.com> wrote:
>
>
>
> > PS. Error msg is "The application failed to initialise properly
> > (0xc0000135)"
>

> > On 7/31/07, Keith Beaumont <beaumon...@gmail.com> wrote:
>
> > > Is this a windows .exe? Don't think so. Can you compile one? If not can
> > > you release the source? Can anybody else then pick this up & do the
> > > necessary?
> > > KB
>

[withbl...@gmail.com]

Hi unknowner,
Appreciated .for your good idea.
If possible compile with VisualStudio6 also and upload with source code for bbtesseract so that it can be re-compiled either with Visual basic6 or Visual basic2005 express along with tesseract2.0,  Is it possible to avoid using dot net2.0?
-74yrsold

On 7/31/07, unknowner <struthe...@googlemail.com> wrote:

[unknowner]

Hi!

uploaded the complete project ( hhttp://groups.google.com/group/tesseract-ocr/files
- > bbTesseract.zip )

I uploaded the complete project for vb2005 express. I don't know if
its possible to compile it with vb6, maybe someone else out there
does?
I think VB2005Ex. requires dot net framework, but as said before maybe
some else knows/finds a way.

On 1 Aug., 08:26, withblessi...@gmail.com wrote:
> Hi unknowner,
> Appreciated .for your good idea.
> If possible compile with VisualStudio6 also and upload with source code for
> bbtesseract so that it can be re-compiled either with Visual basic6 or
> Visual basic2005 express along with tesseract2.0, Is it possible to avoid
> using dot net2.0?
> -74yrsold
>

[Keith Beaumont]

Hello David,

That explains why the .exe is so small !! Yes please for the source. I don't have a VB compiler, but it would be interesting to see the logic. How did you work this out? By looking at the c-code or is there a description of how the program works? Is there a debug option on the tesseract command line? That might be a good starting point for me.

 

I have some very specific requirements in terms of page size & line positioning. If I could change the code to know the page size and point to a position (col/row) on the page where the text starts, then I think that could make the whole process simpler/easier for training. The documents are 18/19 th century handwritten on lined paper about the size of (or bigger) than foolscap. Clearly very difficult for the software, but your box logic might be very useful.

I'm assuming your box is a grid placed over the image. Is that correct or am I on the wrong track?

You may be interested to know that there are thousands of pages of these documents. They are large bound volumes of land deeds (1707-1850 ish) held in a registry near my home in England. At first glance they are unreadable by a modern day person. Think the script is "court hand" or something like it. I would be very excited/overjoyed if tesseract could be trained to interpret this. It might be a major achievement for posterity !! (Monty Python jokes come to mind!)

 

By the way, thanks for the pointer to the PDF.

Where are you? I assume your in the US. And theRaySmith, are you American too?

KB

 

On 7/31/07, unknowner <struthe...@googlemail.com> wrote:

hi keith,
it is a windows exe and it requires dot net framework ( ->thats your
error 0xc0000135.) Sorry for that.

This tool was made with VisualBasic 2005 Express, if you still want
the Source will  upload it.




On 31 Jul., 20:11, "Keith Beaumont" <beaumon...@gmail.com > wrote:
> PPS. This couldn't be because I don't have dot net 2.0?
>
> On 7/31/07, Keith Beaumont <beaumon...@gmail.com> wrote:
>
>
>
> > PS. Error msg is "The application failed to initialise properly
> > (0xc0000135)"
>
> > On 7/31/07, Keith Beaumont <beaumon...@gmail.com > wrote:
>
> > > Is this  a windows .exe? Don't think so. Can you compile one? If not can
> > > you release the source? Can anybody else then pick this up & do the
> > > necessary?
> > > KB
>
> > >  On 7/31/07, unknowner <struthers.da...@googlemail.com > wrote:
>
> > > > Hi guys,
> > > > if your going to train an new language this tool might help you a bit.
> > > > It shows the generated boxefiles over the picture.
>
> > > > you need:
> > > > - dot net 2.0
> > > > - picture as uncompressed tif

> > > > - boxfile ( tesseract.exe train.tif train nobatch makebox) in the same

[unknowner]

Hi Keith,

first off all, im American (just born there) but i live in Germany and
i am more German than American :)

i try to explain how this tool works

1 . you need one scan/pic, must be an uncompressed TIF
2. use tesseract to create the boxfile (.txt), both pic and text MUST
have the same name eg. train1.tif train1.txt
3. start bbTesseract (by the way bb=boxbuilder :-\ ) and open the txt
file
4. press draw boxes to see what tesseract found, now you can click on
a letter an see what tesseract made of it.
- you can cange the letter, delete a "letter" as sometimes a no so
good scan has same spots or lines that are marked as letter like "l"
or "."
- you can also add new letters, but you currently have to set all
values manually

!!! Press Save changes after changing box size or letters !!!

5. if you checked your document completely press "Save Box". Now you
have all files needed to continue training.

David

> > > > > > - boxfile (tesseract.exe train.tif train nobatch makebox) in the

[unknowner]

I'm working on the tool to let you set an region were tesseract should
read.
im currently trying to get the dll working, so all steps needed can be
done with bbT.

Every Box is simply draw over the pic, thats why you should not have a
page with about 10.000 letters for training, might take some time :)

PS: Forgot to say: i LOVE Monty Phyton

[Keith Beaumont]

David,

Glad your in Europe. Makes it easier to have real-time email conversations.

Where are you in Germany? I'm in Leeds in the north of England.

bbTA sounds really good. Now I have to get dot net framework. Hope I don't need the VB compiler for this. Do I need a specific version?

How did you get the logic for this? What about my ideas over page-size/positioning.

If your too busy working, please reply later.
KB
 

> > > > > > - both files need to have the same name ( train.tif train.txt )

[Keith Beaumont]

Your dot NET Framework link gives:

"The download you requested is unavailable. If you continue to see this message when trying to access this download, go to the "Search for a Download" area on the Download Center home page."

 

[Keith Beaumont]

Sorry, found another link now!!

[unknowner]

i think selecting a specific regin for tesseract might be easy. i
create a new pic that contains only the region and let tesseract read
that pic for you.

the logic? really easy, i just parsing the complete txt file with the
single letters and box coords and convert them to x and y on the pic
(like pic_maxHeight - box_top = y / y - box_bottom = box_height)

for dot net do a search on google for "Microsoft .NET Framework
Version 2.0", that might work

do you already have a scan of one of the documents

regards david

i'm near Mannheim, South of Frankfurt

On 1 Aug., 11:49, "Keith Beaumont" <beaumon...@gmail.com> wrote:
> Your dot NET Framework link gives:
> "The download you requested is unavailable. If you continue to see this
> message when trying to access this download, go to the "Search for a
> Download" area on the Download Center home page."
>

> On 8/1/07, Keith Beaumont <beaumon...@gmail.com> wrote:
>
>
>
> > David,
> > Glad your in Europe. Makes it easier to have real-time email
> > conversations.
> > Where are you in Germany? I'm in Leeds in the north of England.
> > bbTA sounds really good. Now I have to get dot net framework. Hope I don't
> > need the VB compiler for this. Do I need a specific version?
> > How did you get the logic for this? What about my ideas over
> > page-size/positioning.
> > If your too busy working, please reply later.
> > KB
>

[Keith Beaumont]

I hadn't realised that the output from training were coordinates. I'd assumed something silly like codes for the characters. Now it's much clearer!!  Why 4 columns ? NO, now it's clear, I think.

It's x1 y1, x2 y2 where 1 is start, 2 is end. x/y axis starts at bottom left of document as in normal graph. What are the units? Pixcels?
I have a sample, but it's not of the deed books, the real target. The sample is a very large image. Just cropped it from the centre & now going to try bbT. Will let you know results.
 

I cycled through your town once. It's on/near the Rhine isn't it? I was on my way to Heidelberg, a slightly more interesting town (sorry!!). Do you work in Mannheim?

 

On 8/1/07, unknowner <struthe...@googlemail.com> wrote:

[unknowner]

Yes, it is near the Rhine! And your true, Heidelberg is more
interesting and much nicer than mannheim. I work in Viernheim, right
next to mannheim and live near Weinheim, less polution and traffic
than mannheim :)

On 1 Aug., 12:28, "Keith Beaumont" <beaumon...@gmail.com> wrote:
> I hadn't realised that the output from training were coordinates. I'd
> assumed something silly like codes for the characters. Now it's much
> clearer!! Why 4 columns ? NO, now it's clear, I think.
> It's x1 y1, x2 y2 where 1 is start, 2 is end. x/y axis starts at bottom left
> of document as in normal graph. What are the units? Pixcels?
> I have a sample, but it's not of the deed books, the real target. The sample
> is a very large image. Just cropped it from the centre & now going to try
> bbT. Will let you know results.
>
> I cycled through your town once. It's on/near the Rhine isn't it? I was on
> my way to Heidelberg, a slightly more interesting town (sorry!!). Do you
> work in Mannheim?
>

[Keith Beaumont]

This is brilliant!! Have a look. Not sure how far we can get with this document. Remember, this is NOT my target document. This is just an image the archive happened to have.

First thing is to split the boxes with more than one character.

 

The word "do" means "the same as above", in case you don't know English so well now.
 

Used: tesseract.exe 0001a.tif 0001a nobatch makebox
 

On 8/1/07, unknowner <struthe...@googlemail.com> wrote:

> > > > > > > > > > > - boxfile ( tesseract.exe train.tif train nobatch

[Keith Beaumont]

On 8/1/07, Keith Beaumont <beaum...@gmail.com> wrote:

[Keith Beaumont]

The original tif is too large for Google. Here is the jpg I started with.

I converted to tif with ReaConverterPro.exe

 

On 8/1/07, Keith Beaumont <beaum...@gmail.com> wrote:

Here's the complete document. Perhaps I need to break it down into clearer units?

I'm about to read the training directions as I don't know how to change the box sizes in such a way that tesseract will know what to do for future documents. Any advice would be welcome.

[unknowner]

As Tesseract did a pretty good job on your first pic, i tried a edit
your txt file. you can save this into 0001a.txt file (backup your
original) an load it ionto bbT. With enough Trainingsheets Tesseract
will do a pretty good job on you documents.

1 174 957 193 984
0 189 961 204 981
d 280 963 303 983
o 303 963 312 981
d 621 962 642 982
o 643 961 650 981
& 662 955 678 980
d 700 962 720 984
o 717 963 724 979
4 806 960 825 982
1 843 963 858 981
,, 885 960 892 968
, 960 965 964 973
1 187 927 200 947
/ 194 925 207 945
d 280 925 300 946
o 300 932 310 942
G 631 915 655 948
r 653 927 663 940
a 663 926 679 939
s 678 926 692 942
s 690 927 704 943
,, 847 926 854 936
. 889 926 893 932
,, 962 925 968 934
/ 184 891 197 910
2 194 890 208 910
_ 224 891 270 895
w 282 891 310 913
_ 329 890 363 895
. 361 890 392 895
. 391 890 411 894
, 410 890 418 895
_ 417 890 460 895
_ 459 890 484 894
. 483 891 558 894
. 563 891 567 895

[Keith Beaumont]

OK. That's good. What tool should I use to break the large image into several smaller ones? Is there a Microsoft tool to do this? I'm VERY excited!!

I'm trying to read http://code.google.com/p/tesseract-ocr/wiki/TrainingTesseract at the moment.

Which file tells tesseract about the box sizes? Is it the box file you just edited or is this re-input into ta & another file comes out?
 

On 8/1/07, unknowner <struthe...@googlemail.com> wrote:

[Keith Beaumont]

Now I've got it!  We use box.train to get the .tr

Then I have to read more!!!

 

On 8/1/07, Keith Beaumont <beaum...@gmail.com> wrote:

[Keith Beaumont]

I'm getting there now - slowly. I used notepad to edit the box file and insert your stuff. Does this produce UTF-8 characters? Do you have to use wordpad for unicharset or other files?

[unknowner]

I used both, they both can save text as unicode,

[Keith Beaumont]

You said:

i think selecting a specific regin for tesseract might be easy. i
create a new pic that contains only the region and let tesseract read
that pic for you.
 

I need to breakdown the large image into smaller units. Which tool should I use?

Tried tacmi pixia, but MsOffice says the saved tif is not recognised & tesseract crashes!!

 

On 8/1/07, unknowner <struthe...@googlemail.com> wrote:

[unknowner]

for now you can use something like http://www.irfanview.net/ , its a
pic viewer and lets you save pics as uncompressed tif

On 1 Aug., 14:51, "Keith Beaumont" <beaumon...@gmail.com> wrote:
> You said:
> i think selecting a specific regin for tesseract might be easy. i
> create a new pic that contains only the region and let tesseract read
> that pic for you.
>
> I need to breakdown the large image into smaller units. Which tool should I
> use?
> Tried tacmi pixia, but MsOffice says the saved tif is not recognised &
> tesseract crashes!!
>

[Keith Beaumont]

Found ImageCut:

http://www.filedudes.com/Image_Cut__Image_splitter_-getfile-22510.html
 

It works very well. You can split into several vertical/horizontal sections in one go. Perfect.

Thanks for all your help so far. Now back to it!!
 

On 8/1/07, unknowner <struthe...@googlemail.com> wrote:

[Keith Beaumont]

Tesseract wouldn't work with tif files from ImageCut. Think there was some compression.

So downloaded irfanview.

So far have 2 tif/box file pairs. Tried to run ta for training, but no luck.
command was: tesseract.exe tithe1.tif junk1 nobatch  box.train
 

Here are the files. Can you help?

 

On 8/1/07, Keith Beaumont <beaum...@gmail.com> wrote:

[Keith Beaumont]

David,

If I rename the output file of: tesseract.exe tiff3.tif tiff3 nobatch makebox

from .txt to .box without editing, it works fine! Also the small file you sent earlier works OK.
 

I produced these files tiff1/2 using bbta.

The process was do a bit of work, save output, rename old .txt file, rename .box to .txt

Use bbta again & so on until all fixes done. Also edited a bit in between with notepad. Are these steps valid?

I'm a bit tired!! Maybe i've mixed something up?

KB
 

[Keith Beaumont]

Got it, you have to save the box files as UTF-8 files in notepad.

I can sleep easy tonight!!

Oh, I'm so tired. My brain hurts!!

Well back to the tesseract grindstone tomorrow.
Thanks again for your help David.

[Keith Beaumont]

Morning David,

Currently editing third file with your bb. What's the minimum number of training files I can get away with in your experience? I'd like to know this is going to work ASAP !!

 

Is it better to make the boxes smaller or larger? The minimum space to define a character maybe chopping a bit off or the maximum which could include a bit of the next character. Do you have a "feel" for this?

 

Have you thought about dragging the edges/corners of the boxes to resize them in bb?

Also maybe drag & drop a bit of the rubbish to add a new box where one is missing?
KB
 

[unknowner]

Hi Keith,

can't tell you how much training files required, but its possible to
use up to 32. I guess more is better.

Been working on bbT and also thought about drag and drop and it looks
like it may work :), i think i need about two days to release the next
version so V0.0000000 ...0004 is on its was :)

On 2 Aug., 10:43, "Keith Beaumont" <beaumon...@gmail.com> wrote:
> Morning David,
> Currently editing third file with your bb. What's the minimum number of
> training files I can get away with in your experience? I'd like to know this
> is going to work ASAP !!
>
> Is it better to make the boxes smaller or larger? The minimum space to
> define a character maybe chopping a bit off or the maximum which could
> include a bit of the next character. Do you have a "feel" for this?
>
> Have you thought about dragging the edges/corners of the boxes to resize
> them in bb?
> Also maybe drag & drop a bit of the rubbish to add a new box where one is
> missing?
> KB
>

> On 8/1/07, Keith Beaumont <beaumon...@gmail.com> wrote:
>
>
>
> > Got it, you have to save the box files as UTF-8 files in notepad.
> > I can sleep easy tonight!!
> > Oh, I'm so tired. My brain hurts!!
> > Well back to the tesseract grindstone tomorrow.
> > Thanks again for your help David.
>

> > On 8/1/07, Keith Beaumont <beaumon...@gmail.com> wrote:
>
> > > David,
> > > If I rename the output file of: tesseract.exe tiff3.tif tiff3 nobatch
> > > makebox
> > > from .txt to .box without editing, it works fine! Also the small file
> > > you sent earlier works OK.
>
> > > I produced these files tiff1/2 using bbta.
> > > The process was do a bit of work, save output, rename old .txt file,
> > > rename .box to .txt
> > > Use bbta again & so on until all fixes done. Also edited a bit in
> > > between with notepad. Are these steps valid?
> > > I'm a bit tired!! Maybe i've mixed something up?
> > > KB
>

> > > On 8/1/07, Keith Beaumont <beaumon...@gmail.com > wrote:
>
> > > > Tesseract wouldn't work with tif files from ImageCut. Think there was
> > > > some compression.
> > > > So downloaded irfanview.
> > > > So far have 2 tif/box file pairs. Tried to run ta for training, but no
> > > > luck.
> > > > command was: tesseract.exe tithe1.tif junk1 nobatch box.train
>
> > > > Here are the files. Can you help?
>

> > > > On 8/1/07, Keith Beaumont <beaumon...@gmail.com > wrote:
>
> > > > > Found ImageCut:
> > > > >http://www.filedudes.com/Image_Cut__Image_splitter_-getfile-22510.html
>
> > > > > It works very well. You can split into several vertical/horizontal
> > > > > sections in one go. Perfect.
> > > > > Thanks for all your help so far. Now back to it!!
>

> > > > > On 8/1/07, unknowner <struthers.da...@googlemail.com > wrote:
>
> > > > > > for now you can use something likehttp://www.irfanview.net/, its

[Keith Beaumont]

David,

HELP ....

All steps performed with 3 training files. Run: tesseract.exe C:\ocr\big\tithe.tif C:\ocr\big\tithe.hope -l enm

log file reads: Unable to load unicharset file C:/tesseract-2.00/tessdata/enm.unicharset

File is in place in UTF-8 format. Any ideas?

 

Please convert 0001.jpg & name as tithe.tif or use one of the smaller tif's from yesterday.
 

[unknowner]

Hi Keith,

i nver had that error, please start new topic, maybe someeles can help

On 2 Aug., 12:38, "Keith Beaumont" <beaumon...@gmail.com> wrote:
> David,

> HELP ....
> All steps performed with 3 training files. Run:

> tesseract.exeC:\ocr\big\tithe.tif C:\ocr\big\tithe.hope -l enm

> log file reads: Unable to load unicharset file C:/tesseract-2.00
> /tessdata/enm.unicharset
> File is in place in UTF-8 format. Any ideas?
>
> Please convert 0001.jpg & name as tithe.tif or use one of the smaller tif's
> from yesterday.
>

> [enm.DangAmbigs]
>
> enm.freq-dawg
> 1KHerunterladen
>
> enm.inttemp
> 238KHerunterladen
>
> [enm.normproto]4
> linear essential -0.250000 0.750000
> linear essential 0.000000 1.000000
> linear essential 0.000000 1.000000
> linear essential 0.000000 1.000000
>
> 1 9
> significant elliptical 1
> 0.312539 0.131284 0.150572 0.103397
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.290268 0.162955 0.197321 0.121021
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.294382 0.145203 0.177977 0.109825
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.322986 0.170473 0.208875 0.150633
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.371301 0.258383 0.268021 0.183187
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.593180 0.218983 0.263038 0.169452
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.611292 0.189785 0.251813 0.137023
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.603855 0.279758 0.267126 0.210174
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.590205 0.297036 0.252068 0.216494
> 0.000100 0.000100 0.000100 0.000100
>
> e 7
> significant elliptical 1
> 0.166719 0.122418 0.112509 0.117842
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.163663 0.093659 0.102489 0.092770
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.228266 0.104660 0.123470 0.113054
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.181539 0.187512 0.161878 0.122033
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.201541 0.194026 0.146079 0.104562
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.207836 0.265588 0.138891 0.200688
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.236132 0.323519 0.119861 0.187207
> 0.000100 0.000100 0.000100 0.000100
>
> l 3
> significant elliptical 1
> 0.415204 0.233662 0.233677 0.141975
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.322588 0.243253 0.181222 0.147935
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.520135 0.306385 0.268884 0.174135
> 0.000100 0.000100 0.000100 0.000100
>
> d 13
> significant elliptical 1
> 0.528414 0.441816 0.254538 0.251238
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.456736 0.502622 0.231320 0.237700
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 8
> 0.314058 0.431626 0.231076 0.240821
> 0.000765 0.000996 0.000251 0.000332
> significant elliptical 1
> 0.358659 0.354384 0.192812 0.204723
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.199494 0.345508 0.164871 0.199934
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.263055 0.322313 0.181719 0.213100
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 2
> 0.289594 0.330290 0.198744 0.219730
> 0.000100 0.000134 0.000100 0.000100
> significant elliptical 2
> 0.254717 0.299265 0.175020 0.162982
> 0.000187 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.258754 0.377284 0.160711 0.177796
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.181813 0.350876 0.213393 0.219889
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.158820 0.376872 0.249377 0.205152
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.131787 0.426549 0.228889 0.242218
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.177547 0.102294 0.129519 0.063750
> 0.000100 0.000100 0.000100 0.000100
>
> A 3
> significant elliptical 1
> 0.226092 0.304464 0.171679 0.194327
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.280890 0.311974 0.152159 0.179834
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.355369 0.449401 0.226852 0.222887
> 0.000100 0.000100 0.000100 0.000100
>
> r 2
> significant elliptical 14
> 0.199583 0.121317 0.117735 0.058545
> 0.001149 0.000770 0.000367 0.000249
> significant elliptical 1
> 0.069453 0.138286 0.042012 0.095710
> 0.000100 0.000100 0.000100 0.000100
>
> a 8
> significant elliptical 2
> 0.235517 0.268561 0.131932 0.170841
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.171533 0.278663 0.130817 0.210706
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.208159 0.348790 0.141102 0.214314
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 2
> 0.166270 0.242089 0.092921 0.147569
> 0.000100 0.000100 0.000100 0.000170
> significant elliptical 1
> 0.145038 0.222962 0.100473 0.145273
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.181097 0.226117 0.110436 0.134336
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.170177 0.192333 0.121736 0.106461
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 5
> 0.117406 0.176338 0.081737 0.105454
> 0.000320 0.000197 0.000140 0.000106
>
> & 5
> significant elliptical 1
> 0.156279 0.244192 0.232996 0.091428
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.253954 0.280720 0.208405 0.114290
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.287129 0.237332 0.218165 0.070943
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.261486 0.341359 0.263355 0.142402
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.377795 0.351098 0.257841 0.145005
> 0.000100 0.000100 0.000100 0.000100
>
> M 1
> significant elliptical 1
> 0.251229 0.406011 0.155065 0.226104
> 0.000100 0.000100 0.000100 0.000100
>
> 2 2
> significant elliptical 1
> 0.294859 0.322741 0.203349 0.130968
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.363308 0.246644 0.185225 0.107226
> 0.000100 0.000100 0.000100 0.000100
>
> o 2
> significant elliptical 14
> 0.447151 0.116454 0.095232 0.075134
> 0.015857 0.000427 0.000191 0.000298
> significant elliptical 1
> 0.237719 0.343533 0.141550 0.236523
> 0.000100 0.000100 0.000100 0.000100
>
> G 3
> significant elliptical 3
> 0.227121 0.434326 0.262072 0.156217
> 0.000596 0.007223 0.000100 0.000100
> significant elliptical 4
> 0.251821 0.583814 0.381243 0.210076
> 0.003237 0.000592 0.001787 0.000637
> significant elliptical 1
> 0.128197 0.276767 0.180804 0.107874
> 0.000100 0.000100 0.000100 0.000100
>
> s 8
> significant elliptical 4
> 0.268567 0.232846 0.162817 0.149573
> 0.000100 0.000100 0.000150 0.000115
> significant elliptical 1
> 0.215375 0.196499 0.150038 0.120459
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.203063 0.187275 0.162802 0.118285
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.248215 0.285642 0.122705 0.193605
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.256769 0.297521 0.114163 0.218819
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.221271 0.512682 0.118799 0.316383
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.139253 1.175681 0.304155 0.313994
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.148099 1.184912 0.279563 0.328478
> 0.000100 0.000100 0.000100 0.000100
>
> 3 1
> significant elliptical 1
> 0.322552 0.358422 0.240355 0.165715
> 0.000100 0.000100 0.000100 0.000100
>
> H 1
> significant elliptical 1
> 0.459131 0.878914 0.278299 0.308716
> 0.000100 0.000100 0.000100 0.000100
>
> u 1
> significant elliptical 1
> 0.212725 0.159948 0.147559 0.112509
> 0.000100 0.000100 0.000100 0.000100
>
> t 1
> significant elliptical 1
> 0.327241 0.174999 0.210061 0.130996
> 0.000100 0.000100 0.000100 0.000100
>
> y 1
> significant elliptical 1
> 0.090702 0.389284 0.256174 0.130484
> 0.000100 0.000100 0.000100 0.000100
>
> 4 3
> significant elliptical 1
> 0.425548 0.345923 0.237301 0.165028
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.301800 0.328058 0.241161 0.171823
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.141516 0.292822 0.215167 0.150440
> 0.000100 0.000100 0.000100 0.000100
>
> n 2
> significant elliptical 1
> 0.210714 0.318404 0.124366 0.287453
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.193068 0.176226 0.077774 0.139203
> 0.000100 0.000100 0.000100 0.000100
>
> 5 1
> significant elliptical 1
> 0.226631 0.224566 0.161513 0.088247
> 0.000100 0.000100 0.000100 0.000100
>
> 0 2
> significant elliptical 1
> 0.414933 0.342729 0.212685 0.168413
> 0.000100 0.000100 0.000100 0.000100
> significant elliptical 1
> 0.240210 0.203185 0.142850 0.118448
> 0.000100 0.000100 0.000100 0.000100
>
> c 1
> significant elliptical 1
> 0.117302 0.115162 0.070441 0.112125
> 0.000100 0.000100 0.000100 0.000100
>
> h 1
> significant elliptical 1
> 0.236586 0.240227 0.154374 0.088623
> 0.000100 0.000100 0.000100 0.000100
>
> 6 2
> significant elliptical 1
> 0.401273 0.324482 0.224927 0.114246
> ...
>
> Erfahren Sie mehr »
>
> [enm.pffmtable]1 70
> e 100
> l 57
> d 82
> A 79
> r 80
> a 81
> & 156
> M 98
> 2 80
> o 90
> G 134
> s 224
> 3 69
> H 117
> u 49
> t 43
> y 73
> 4 71
> n 84
> 5 70
> 0 68
> c 61
> h 72
> 6 72
> 7 53
> i 81
> 8 73
> 9 80
> b 64
> L 66
>
> enm.unicharset
> 1KHerunterladen
>
> [enm.user-words]
>
> enm.word-dawg
> 1KHerunterladen
>
> 0001.jpg
> 1003KAnzeigenHerunterladen

[Keith Beaumont]

David,

Have you tried to run with the files I sent? Does it fail for you? If it works, please mail output file.

Is there anything in the file it doesn't like?

The file starts with the number of characters. That must be normal?

I tried deleting the funnies like null , but no joy.
KB
 

On 8/2/07, unknowner <struthe...@googlemail.com> wrote:

[Keith Beaumont]

Think i've got it. Look at eng.unicharset with notepad. Probably needs a unix file format!!

Im now trying to edit with wordpad from eng version. Think it will maintain correct line-throw characters.

Have you built your unicharset yet?

 

[Keith Beaumont]

Can't edit this file with wordpad !! Changes the form-feed chars !!

So copy eng. to my enm. & get new message in log:

Error: 32 classes in inttemp while unicharset contains 112 unichars.

 

So do you know of a unix editor for windows?

If I can remember the ascii codes, I might be able to do it with awk.

Now need an ascii viewer. Think I remember there are ascii(hex) editors somehere.
Any ideas?
 

[beaum...@gmail.com]

David,
Still have my problems with unicharset. See issue 49. I was on the
wrong track with linefeed characters.
Now looking at the other files normproto, etc. to try & count
characters and get rid of the 2 that are causing the problem.
But inttemp is not a readable file for mere mortals. Do you know if
there is a debug option for mftraining?
KB

On Aug 2, 12:49 pm, "Keith Beaumont" <beaumon...@gmail.com> wrote:
> Can't edit this file with wordpad !! Changes the form-feed chars !!
> So copy eng. to my enm. & get new message in log:
> Error: 32 classes in inttemp while unicharset contains 112 unichars.
>
> So do you know of a unix editor for windows?
> If I can remember the ascii codes, I might be able to do it with awk.
> Now need an ascii viewer. Think I remember there are ascii(hex) editors
> somehere.
> Any ideas?
>

> On 8/2/07, Keith Beaumont <beaumon...@gmail.com> wrote:
>
>
>
>
>
> > Think i've got it. Look at eng.unicharset with notepad. Probably needs a
> > unix file format!!
> > Im now trying to edit with wordpad from eng version. Think it will
> > maintain correct line-throw characters.
> > Have you built your unicharset yet?
>

> > On 8/2/07, Keith Beaumont <beaumon...@gmail.com> wrote:
>
> > > David,
> > > Have you tried to run with the files I sent? Does it fail for you? If it
> > > works, please mail output file.
> > > Is there anything in the file it doesn't like?
> > > The file starts with the number of characters. That must be normal?
> > > I tried deleting the funnies like null , but no joy.
> > > KB
>

> ...
>
> read more »- Hide quoted text -
>
> - Show quoted text -

[unknowner]

I'm currently a bit busy, but i will check your files today in the
evening, i let you know what error messages i get :)

David

beaum...@gmail.com schrieb:

[beaum...@gmail.com]

Forget it! Finally solved it. See my comment at end of issue 49.
Now my OCR output is NOT very good with 3 training files. Trying to
discover if even 1 character came out !!
Whatever, can only carry on training or give it up !! How many
training files before I'll know if it can't be done?
This is a rhetorical question (sorry!)

On Aug 3, 12:04 pm, unknowner <struthers.da...@googlemail.com> wrote:
> I'm currently a bit busy, but i will check your files today in the
> evening, i let you know what error messages i get :)
>
> David
>

> beaumon...@gmail.com schrieb:

[beaum...@gmail.com]

David,
Back with bb now. Suggestions:
1 Show name of box file being edited in title bar.
2 Clear edit box & pos wheels when new file loaded.
3 Include *.box in file open dialog - can then re-edit saved box file
(It's easier to delete rubbish with notepad & then try again).
Or Extra save button. Save as .txt. (Might then have to re-name
original .txt file)

On Aug 3, 12:52 pm, "beaumon...@gmail.com" <beaumon...@gmail.com>
wrote:

[Keith Beaumont]

PS.

Can you make the minimum window size smaller?

It currently fills about 80% of the screen & won't reduce further.

 

[unknowner]

Hi Keith,

included all your ideas, as they are quiet good. Still got some bugs
to fix and going to release V0.1 :) by tomorrow i hope.

To everybody: if you got some ideas, don't be shy and post them
here :)

David

On 3 Aug., 17:00, "Keith Beaumont" <beaumon...@gmail.com> wrote:
> PS.
> Can you make the minimum window size smaller?
> It currently fills about 80% of the screen & won't reduce further.
>

> ...
>
> Erfahren Sie mehr »

[unknowner]

Hi out there,

finally made it! V0.1b ;)
You can download it here: http://groups.google.com/group/tesseract-ocr/files
>> bbTesseractV0_1b.gz

It still might have some errors, don't mind and post everything right
here. I will have a look at it as soon as possible.
If you shave some ideas do the same.

regards
David

[unknowner]

Forgot to say if want get the source, i can send it to you. I will
upload it a bit later, as i'm currently cleaning up the code

[withbl...@gmail.com]

Hi David,
You have done excellent job - very convenient, beautiful  and user friendly also.
Edit box can  be improved  to accommodate for  non-english  languages at least  up-to 4 characters - so that even single character  or more characters can be edited  . Create beta version and upload for beta testing/feedback by others to you.  

One More suggestion, It is felt that bbt can also be improved further  so that  all  commandlines  executed  in  DOS,  can be  incorporated in  GUI -bbt, thereby avioding usage of  "commandprompt". and output will be usual format.
Advantage of GUI is  UTF-8 supports whereas "comandprompt" does not support UTF-8 except English.

-74yrsold-

[David Struthers]

Hi Quan,

im still working on the next release, but as my last pc broke down and
work is still quiet busy it will take some time.

I attached the last version, which ist still alpha but works quit good.
Maybe this works better for you and solves some issue. I included the (i
hope) complete project for VB2005 for that one.

Just let me know if this worked for you.

regards
David

nguyenq schrieb:
> Hi David,
>
> Thank you for creating an intuitive tool for editing the box files.
>
> There is a bug that crashes the program every time, however,
> preventing me from utilizing the tool to produce good box files for
> Vietnamese language. If there are two adjacent boxes and I delete one,
> select the other and expand it, via the Right NumericUpDown control
> for example, to cover the area vacated by the deleted box, the program
> goes into an unstable state. As I select another boxed character, an
> erroneous, duplicated character will appear in the area below while
> removing the rectangular box around the selected character. Repeated
> selection of other boxes will eventually generate an
> ArgumentException, due to passing in a negative value to the Bitmap
> constructor inside the Form_bbTeseract.crop method.
>
> There is also a problem opening the Form_bbTeseract form. Visual
> Studio 2005 IDE would report "one or more errors encountered while
> loading the designer".
>
> The project seems to be missing the left & right arrow image files.
>
> Hope you will fix them soon. Thank you.
>
> Regards,
> Quan Nguyen
>
>
>

[Basu]

Thanks David for the new version of bbt.
Its working fine..
Specially with multi-character boxes.
Lot of improvement in that area in cmparison to your last available
version.
bbt appears to be more robust this time.
I have two immediate suggestions..
1. Can you include a bmp file support (not much effort I suppose)? May
be somebody like me would be interested. Though its not space
efficient but popular due to its simplicity. Tesseract supports bmp as
well.
2. Some unhandled exceptions for invalid (box)*.txt files (even if
*.tif file exists in same name). It would be great if you can show
some messege instead of quitting application.

and thanks for your implicit update on 7z extension.
I had no idea about it.
For all such ignorants like me, plz refer http://www.7-zip.org/ and
download the 7z extractor.
Thanks.
Basu

On Nov 7, 2:09 am, David Struthers <struthers.da...@googlemail.com>
wrote:

> bbt.7z
> 896KDownload
>
> bbTesseract.7z
> 34KDownload- Hide quoted text -

[unknowner]

Can you please send me the box file and tif is possible, so i can
check whats going on there.

bmp support is going to be in the next version, currently i got a
little bit time to work on bbT

David

[unknowner]

I uploaded the exe & source to the files page for everyone interested
(0.2a (or b) :) )

> > > - Show quoted text -- Zitierten Text ausblenden -
>
> - Zitierten Text anzeigen -

[Basu]

Hi David,
actually, I have "filename.txt" as my input data in digital pen format
Tess generates "filename.box.txt" for me
the image is "filename.bmp" or "filename.tif"
from bbt if I try to open "filename.box.txt" it gives an obvious error
messege that "filename.box.tif" not found
but if I try to open (by mistake) "filename.txt" (which is not
actually a box file) bbt gives unhandled error and quits.
It may be nice, if the ".txt" file format is not ok, a suitable
messege is displayed. like "invalid text file".
I just wanted to give this suggestion..but its a trivial problem.

I have unfortunately come accross some more bugs.
1. its difficult to edit the labels. one can only edit the character-
codes and save. if you change the characters, its not saving.but still
I could edit the boxes by editing the characters codes.
2. if you click on some box, and change its coordinates, the edited-
box often disappears from the image (shifted to the right). This
problem is preety consistent, I think you might have encountered as
well. But this is only display problem. The coordinate-changes are
stored properly in boxfile. So to see the correct boxes, I need to
reopen the boxfile.
3. even sometimes, if I just click on some box, its shape changes.
but only in display (no change in actual box file).

I am uploading a basu.zip file. Please click on the box of 'p' and try
to change the bottom coordinate.
In my machine, the box of 'p' disappears and can be found at the
rightmost part of image.

But still I like bbt very much, and great work anyway!
cheers
Basu

> > - Zitierten Text anzeigen -- Hide quoted text -

[nguyenq]

Thanks, David. It still crashes a lot, such as when adding a new box.
There are still problems with boxes disappearing. Due to the big size
of Vietnamese alphabets' training samples, it takes a long time to
load the box file. Keep reloading the file everytime the program
crashes is impractical. I decided to wait for the next stable release.

Regards,
Quan

On Nov 6, 11:09 am, David Struthers <struthers.da...@googlemail.com>
wrote:

> bbt.7z
> 896KDownload
>
> bbTesseract.7z
> 34KDownload

[nguyenq]

On Nov 7, 6:49 pm, Basu <bsubha...@gmail.com> wrote:
> I have unfortunately come accross some more bugs.
> 1. its difficult to edit the labels. one can only edit the character-
> codes and save. if you change the characters, its not saving.but still
> I could edit the boxes by editing the characters codes.

May I suggest to commit the change, but not yet save the file, when
the focus leaves the edit box, such as when the user click on another
box to edit?

> 2. if you click on some box, and change its coordinates, the edited-
> box often disappears from the image (shifted to the right). This
> problem is preety consistent, I think you might have encountered as
> well. But this is only display problem. The coordinate-changes are
> stored properly in boxfile. So to see the correct boxes, I need to
> reopen the boxfile.

I experience the same problem.