I really love your idea of making this package open source!  I'm
trying to experiment with your layout analysis system, but I'm stuck
on some really basic steps and I could use a little help.  I'm working
on a Microsoft Vista 64-bit system.  I tried simple strategies for
porting to Visual Studio 2005, but quickly realized that I would have
to do a significant amount of rewriting to make this code portable to
both Linux and Windows.  So I switched to cygwin and the GNU toolset.
After resolving a few issues I now have a fully compiled OCRopus
project with tesseract and iulib (but not openFST, which has some
issues with cygwin that I have not yet resolved).  This runs just fine
on the alice_1.png file, and tesseract runs well enough on its test
set.

So now I'm trying to run much more difficult files through OCRopus
(large newspaper scans), which mostly fail with obscure warnings
printed from the code (too many columns, various calculations not
yielding plausible values, etc.).  I'd like to debug the code, setting
breakpoints, modifying the code, etc.  I've tried using Eclipse
running in conjunction with cygwin as a development environment, but
this hits a variety of bizarre errors that I suspect have to do with
mixing Windows paths that start with a drive letter with unix-style
paths.  Now I'm trying to use ddd running on XTerm, with gdb
underneath.  This functions correctly and I can run "ocroscript
recognize data/pages/alice_1.png" correctly, and I can set breakpoints
in ocroscript.cc and ocrtoplevel.cc, I can single step, etc.

But I'm having trouble with the next step.  The recognize.lua script
works correctly, but many of the other scripts appear to be written
for a different LUA interface and will not run with ocroscript.  I can
successfully augment the recognize.lua script by adding commands to
write binary and gray level files (though writing the output file from
the segmenter does not produce the type of zoned image that I might
have expected).  Following the recognize.lua script sequence, I try to
set breakpoints in files like ocr-binarize-sauvola.cc or even in
image.cc (in functions that appear to implement write_image_gray(),
for example, after making sure that I call this function repeatedly
from recognize.lua), but these breakpoints are never reached.  Perhaps
gdb can't set breakpoints in the libocropus.a library that is linked
to ocroscript?

I've also considered compiling some of the top level C++ programs such
as main-ocr-binarize-sauvola.cc, but there aren't any Makefiles that
make this easy.  I can write these Makefiles, but then I'm puzzled as
to precisely which routines I am supposed to be calling and in what
order.  It seems like I ought to be following the recognize.lua
script, but there are multiple possibilities for each phase of
processing and I can't find clear guidance on how to proceed.  The
PowerPoint slide presentations are really interesting and informative,
but many of the lua scripts that appear there either don't exist or
don't work.

I feel like I've missed something really basic and simple.  I know
that you are all working on Ubuntu and I will switch over to that
system if I have to, but what development tools are you using that
make debugging and compiling possible with this code?  Is there some
top level main program that I should be looking at to see how to call
the library functions, and is there some documentation that I've
missed?  I would appreciate any help anyone on this forum can give me.


--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"ocropus" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/ocropus?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to