topic Re: Nuxeo-Platform-OCR Question in Nuxeo Forum

Nuxeo-Platform-OCR Question

Soni_ — Mon, 12 Dec 2011 17:24:09 GMT

Hi:

I'm trying to install 'Nuxeo-platform-ocr' (https://github.com/nuxeo/nuxeo-platform-ocr) , but I do not know where to locate the file 'content_in_doc', so that Nuxeo can use to analyze.

I have followed this manual https://github.com/nuxeo/nuxeo-platform-ocr, but not clear where to locate.

I'm using Ubuntu 10.11 + Tesseract + 3 + Nuxeo Olena (scribe)

Could you tell me where I locate the file 'content_in_doc'?

Thanks, and regards.

Re: Nuxeo-Platform-OCR Question

Olivier_Grisel — Wed, 28 Dec 2011 18:56:50 GMT

I just tried to build against the latest stable version (2.0) of Olena and it seems to work fine. I have updated the README.md of nuxeo-platform-ocr to point to the right source archive.

Beware that the build of olena is has several steps and 2 calls to make in 2 separate folders (the build root and the scribo/src subfolder):

$ wget http://www.lrde.epita.fr/dload/olena/2.0/olena-2.0.tar.bz2
$ tar jxvf olena-*.tar.bz2
$ cd olena-2.0/
$ mkdir _build
$ cd _build
$ ../configure && make
$ cd scribo/src
$ make

The scribo/src should then hold the content_in_doc binary. If not check any error messages in the output the build. Maybe your are missing the development headers for tesseract? Have you installed tesseract 3 from the source tarball and installed it system-wide using sudo make install?

Re: Nuxeo-Platform-OCR Question

rbahntje_Bahntj — Tue, 03 Jan 2012 05:00:30 GMT

I ve compiled Olena 1.0 with Tesseract 3.0 with no problem

Re: Nuxeo-Platform-OCR Question

Olivier_Grisel — Tue, 03 Jan 2012 14:32:37 GMT

As written in the README.md file and as I already answered you have to run make in the $SOURCE_ROOT/_build/scribo/src folder as well and the content_in_doc binary will be created there too.

Re: Nuxeo-Platform-OCR Question

rbahntje_Bahntj — Wed, 04 Jan 2012 05:57:25 GMT

I am running make inside $SOURCE_ROOT/_build/scribo/src folder

Re: Nuxeo-Platform-OCR Question

Olivier_Grisel — Thu, 05 Jan 2012 10:23:16 GMT

I just tried from scratch in a new empty folder from the original tarball and the content_in_doc related lines in the Makefile are not commented out and the binary is built successfully. I suspect that in your case the configure script did not detect some missing dependency

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Fri, 06 Jan 2012 16:49:28 GMT

Right now I'm trying to compile Olena/content_in_doc on Debian Squeeze. I had to install the following packages to make content_in_doc enabled in Makefiles

Re: Nuxeo-Platform-OCR Question

Olivier_Grisel — Mon, 09 Jan 2012 12:11:09 GMT

In my case I built tesseract 3 from the source tarball (as not yet available in ubuntu, I don't know for debian). tesseract 3 gives much better results than tesseract 2 in practice.

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Tue, 10 Jan 2012 10:01:31 GMT

Here I did it using Squeeze's own Tesseract.

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Thu, 09 Feb 2012 16:34:09 GMT

Yet another try. Did it by using (hand-compiled) libleptonica and libtesseract (3). Apparently, Olena 2 only detects the latter when it's compiled "--with-multiple-libraries" (so that it has libtesseract_api.so and so on, and not just libtesseract.so).

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Thu, 09 Feb 2012 18:08:55 GMT

Ok, finally managed to get every piece together (using Olena's git repository instead of release package, and still patching here and there).

First time I imported an image, I had an error about Tesseract being unable to find language data. Right (btw : how do we specify Nuxeo what language it should use to apply OCR?). Then I added the language data, and now I don't have any information about OCR anymore, this is perfectly silent. But no annotations are created.

The only thing that could be related is :

2012-02-09 17:02:36,993 WARN [it.tidalwave.image.java2d.ImplementationFactoryJ2D] JAI not available: java.lang.ClassNotFoundException: javax.media.jai.PlanarImage

Any idea?

Re: Nuxeo-Platform-OCR Question

rbahntje_Bahntj — Fri, 24 Feb 2012 13:32:38 GMT

Finally I made a fresh install from scratch in an Oracle ELinux 5U7 and I can get the content_in_doc binary (I was missing the GDCM2 library) but now I am having the same issue than OlivierM, when I upload an image the server.log show this message

Re: Nuxeo-Platform-OCR Question

rbahntje_Bahntj — Fri, 24 Feb 2012 15:03:57 GMT

I ve installed the JAI package ( http://www.oracle.com/technetwork/java/javasebusiness/downloads/java-archive-downloads-java-client-419417.html#7341-JAI-1.1.2-oth-JPR ), and copy the jai_codec.jar, jai_core.jar and mlibwrapper_jai.jar in mi $NUXEOP_HOME/nxserver/lib

Now I does not get any error messages anymore, but nothing happens when I upload an image file to Nuxeo

How can I debug what is happenning?

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Mon, 27 Feb 2012 09:57:44 GMT

Same here. The JAI warnings disappeared (thanks for the hint!), but nothing is happening.

Re: Nuxeo-Platform-OCR Question

rbahntje_Bahntj — Thu, 22 Mar 2012 18:36:23 GMT

Oliver, did you find a solution?

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Fri, 23 Mar 2012 12:00:01 GMT

Sadly no, I'm still stuck on this, and without time to investigate it further for now.

Re: Nuxeo-Platform-OCR Question

rbahntje_Bahntj — Sun, 01 Apr 2012 01:30:17 GMT

Oliver

The content_in_doc command is working fine. I try to convert an image from the commands lines and it works.

When I upload an image to Nuxeo, I can see a process like this running:

root 25994 25991 97 19:25 pts/0 00:00:15 content_in_doc /opt/nuxeo-cap-5.5-tomcat/tmp/cmdLineBasedConverter22108.jpg /opt/nuxeo-cap-5.5-tomcat/tmp/ocr_olena_1333236340089.xml

And the file ocr_olena_xxxxxxx.xml is created under $NUXEO_HOME/tmp

But..... no annotations are generated in the document in Nuxeo I will try to recompile all again

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Mon, 02 Apr 2012 10:37:25 GMT

Thanks to you, I just discovered ocr_olena_XX.xml files are also created in my tmp directory. Good to know.

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Mon, 02 Apr 2012 11:28:00 GMT

Ok, just a little thing

Re: Nuxeo-Platform-OCR Question

OlivierM_ — Mon, 02 Apr 2012 17:23:58 GMT

I tried to modify the UserPrincipal to an existing user, and the baseURL to my server's, but it doesn't work any better.