Running OSD support of Tesseract OCR

[Chirag Jain]

Dear Sir,

I download tesseract-ocr-3.01-win32-portable.zip from the following
link
http://code.google.com/p/tesseract-ocr/downloads/list

I am finding it difficult to use OSD feature of Tesseract OCR.

I used the following command:

tesseract.exe phototest.tif test_osd -psm 0

It generated test_osd.txt as an output but it was an empty file.

Kindly let me know if I am missing some step.

Thanks and Regards,
Chirag Jain

[Sriranga(78yrs)]

When tested I got "error during the process" displayed in the cmd.

--
You received this message because you are subscribed to the Google
Groups "tesseract-ocr" group.
To post to this group, send email to tesser...@googlegroups.com
To unsubscribe from this group, send email to
tesseract-oc...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/tesseract-ocr?hl=en

[Sriranga(78yrs)]

also got empty text.

[koray]

OSD returns emty text when I tried. Can anyone please clarify if
this is a bug or I m doing things wrong?

[Chirag]

Hi all,

I was able to successfully test orientation detection (after stepping though the code) for various scripts using following commands:

English: tesseract.exe  english_doc.tif  test_osd -l eng -psm 0 

Japanese: tesseract.exe  japanese_doc.tif  test_osd -l jpn -psm 0

Korean: tesseract.exe  korean_doc.tif  test_osd -l kor -psm 0

In these cases, the executable search for eng.traineddata, jpn.traineddata and kor.traineddata respectively along with osd.traineddata.

The performance is really good.

However, it seems like Tesseract is detecting orientation given script.

If I run the executable as following: 

Japanese: tesseract.exe  japanese_doc.tif  test_osd  -psm 0

Korean: tesseract.exe  korean_doc.tif  test_osd  -psm 0

The results are not good. It seems like script detection is not robust.

Am I missing some step? Kindly clarify. 

Regards,
Chirag

On Sat, Mar 3, 2012 at 7:12 PM, koray <uca...@ug.bilkent.edu.tr> wrote:

 OSD returns emty text when I tried. Can anyone please clarify if
this is a bug or I m doing things wrong?

[Sriranga(78yrsold)]

 I noticed "-l lang" before "-psm 0" is missing in your commandline. In the absence of "-l lang" tesseract will always  assume as "-l eng".

extract of help is reproduced below:

M:\>tesseract.exe -h
Usage:tesseract.exe imagename outputbase [-l lang] [-psm pagesegmode] [configfil
e...]
pagesegmode values are:
0 = Orientation and script detection (OSD) only.
1 = Automatic page segmentation with OSD.
2 = Automatic page segmentation, but no OSD, or OCR
3 = Fully automatic page segmentation, but no OSD. (Default)
4 = Assume a single column of text of variable sizes.
5 = Assume a single uniform block of vertically aligned text.
6 = Assume a single uniform block of text.
7 = Treat the image as a single text line.
8 = Treat the image as a single word.
9 = Treat the image as a single word in a circle.
10 = Treat the image as a single character.
-l lang and/or -psm pagesegmode must occur before anyconfigfile.

[Sriranga(78yrsold)]

one more important - please test again as follows:
1st test:tesseract.exe  japanese_doc.tif  test_osd -l jpn -psm 3
2nd test:tesseract.exe  japanese_doc.tif  test_osd         -psm 3
Please check the output text files "test_osd"  - you will find difference in script between two.

[Chirag]

Thanks Sriranga for the response.

If I am specifying the script, it is just orientation detection of that particular script.

The script detection is still a question mark. 

Regards,

Chirag

On Wed, Mar 14, 2012 at 3:51 PM, Sriranga(78yrsold) <withbl...@gmail.com> wrote:

[Chirag]

Just to add one point:

By chance is it assuming it to be English if I don't mention -l <lang>?

In any case, what option/argument should be given for detecting script?

Regards,

Chirag

On Wed, Mar 14, 2012 at 3:22 PM, Chirag <chir...@gmail.com> wrote:

[Chirag]

With -psm 3, I got non-empty files (test_osd.txt)  which were empty with -psm 0. This is true for both with/without -l options.

However, the results of detectOS is same for both -psm [0/3] option for any of with/without -l options.

Please note that I have modified the code slightly to call detectOS separately, which has been doing a good job for orientation detection given script.  I am struggling to detect the script of the input document. 

Regards,
Chirag

[Sriranga(78yrsold)]

I tested using my own lang.tif as follows:
 1) using with -l option -psm 3 ->pl see attached testtif-osd.txt. (non-english)
  2)using without -l option -psm3 ->pl see attached 2testtif-osd.txt. (in English)
In both cases there are no empty output but in different lang

Extract of cmd reproduced below, if used -psm 0
M:\>tesseract.exe test.tif 2testtif-osd    -psm 0  
Tesseract Open Source OCR Engine v3.02 with Leptonica
Error during processing.

M:\>tesseract.exe test.tif 2testtif-osd -l k27   -psm 0
Tesseract Open Source OCR Engine v3.02 with Leptonica
Error during processing.

[Chirag]

Thanks Sriranga for the response.

I was able to perform automatic orientation detection along with page segmentation but I had to supply the script information in the argument for Non English scripts.

I still could not perform automatic script detection. 

Regards
Chirag

2012/3/14 Sriranga(78yrsold) <withbl...@gmail.com>

[Sriranga(78yrsold)]

it would be better to furnish version of tesseract-ocr used and also type of OS used by you.
also upload your used image file for recreating your problem.

[zdenko podobny]

Hi all,

you will not get OSD (Orientation and script detection) output information with tesseract executable. At the moment tesseract provide (save) only ocr result. Somebody could consider help (tesseract --help) misleading because it enumerate all possible page segmentation modes. I think that starting with other than 0 psm would also trigger question...

OSD info you get via API. Here is simple snippet:

    inputfile = "/usr/src/tesseract-3.02/eurotextUpsideDown.png";

    image = pixRead(inputfile);

    api->Init("/usr/src/tesseract-3.02/", "eng");

    api->SetPageSegMode(tesseract::PSM_AUTO_OSD);

    api->SetImage(image);

    api->Recognize(0);

    tesseract::PageIterator* it =  api->AnalyseLayout();

    tesseract::Orientation orientation;

    tesseract::WritingDirection direction;

    tesseract::TextlineOrder order;

    float deskew_angle;

    it->Orientation(&orientation, &direction, &order, &deskew_angle);

    printf("Orientation: %d;\nWritingDirection: %d\nTextlineOrder: %d\n" \

           "Deskew angle: %.4f\n",

           orientation, direction, order, deskew_angle);

In output you can see these information(eurotextUpsideDown.png is 180 degree rotated image eurotext.tif):

Orientation: 2;

WritingDirection: 0

TextlineOrder: 2

Deskew angle: -0.0038

This means that page is upside down (Orientation)[1], text is written from left to right (WritingDirection)[2], lines are ordered from top to down (TextlineOrder)[3], and there is small deskew angle[4].

As you can see attributes WritingDirection, TextlineOrder, Deskew do not reflect that page is upside down, so you will get the same result for eurotext.tif.

[1] http://code.google.com/p/tesseract-ocr/source/browse/trunk/ccstruct/publictypes.h?r=716#81

[2] http://code.google.com/p/tesseract-ocr/source/browse/trunk/ccstruct/publictypes.h?r=716#111

[3] http://code.google.com/p/tesseract-ocr/source/browse/trunk/ccstruct/publictypes.h?r=716#125

[4] http://code.google.com/p/tesseract-ocr/source/browse/trunk/ccmain/pageiterator.h?r=716#239

--

Zdenko

On Tue, Nov 13, 2012 at 10:26 AM, Alex <miller.j...@gmail.com> wrote:

Hello,Chirag.

I am also trying to find a way to detect the script of the input document.

Kindly let me know if  you have some progress.

Thanks and Regards,
Alex



在 2012年3月14日星期三UTC+8下午7时17分29秒，Chirag Jain写道：

[Mikael Egibyan]

I am using /MT lib binaries and your code crashes on

it->Orientation(&orientation, &direction, &order, &deskew_angle);

Any idea what is wrong?

Thanks,
Mikayel

[YounesC]

hi

I use tesseract 3.02

when I use mode PSM_AUTO_OSD,

the orientation finder works just fine:

      tesseract::PageIterator* it = api->AnalyseLayout();

    tesseract::Orientation orientation;

    tesseract::WritingDirection direction;

    tesseract::TextlineOrder order;

    float deskew_angle;

    it->Orientation(&orientation, &direction, &order, &deskew_angle);

    printf("Orientation: %d;\nWritingDirection: %d\nTextlineOrder: %d\n" \

           "Deskew angle: %.4f\n",

           orientation, direction, order, deskew_angle);

When I use PSM_OSD_ONLY, the orientation finder doesn't work.

any ideas?

do I have to upgrade to Tesseract 3.04?