conjecture
An Extensible Optical Character Recognition Framework

Requirements

TODO

Installation

The Conjecture environment currently requires the following process to install:

  1. Obtain the source:
    • From the latest stable release, available as Conjecture-current.tgz (older releases are also available).
      % cd <somedir>
      % wget http://www.holst.ca/conjecture/release/Conjecture-current.tgz
      % ln -s Conjecture-<version> Conjecture
      
    • From the SVN repository:
      % cd <somedir>
      % svn checkout --username <username> 
                     --password <password> 
                     svn://www.holst.ca/Conjecture
      % export CONJECTURE=<somedir>/Conjecture
      
  2. Update environment variables:
    • Set CONJECTURE to <somedir>/Conjecture
    • Add $CONJECTURE/install/bin to your PATH environment variable
    • Add $CONJECTURE/install/lib/perl to your PERLLIB environment variable
    • Add $CONJECTURE/install/lib to your LD_LIBRARY_PATH environment variable

    The above environment variables can be automatically set by using the special $CONJECTURE/config/conjecture.{sh,csh,bat} scripts. For example, if you are using bash, you could add the following line to your .bashrc or .profile file:

       source <somedir>/Conjecture/config/conjecture.bash <somedir>/Conjecture
       
    and it will properly initialize CONJECTURE and update all required environment variables.

    IMPORTANT: Currently, if you do not set these environment variables, everything from initial installation to the execution of the scripts and exectuables will fail miserably. Yes, we will want to remove this burden from the individual installing the code - something to be worked on as Conjecture evolves.

  3. Compile the source code:
    % cd $CONJECTURE
    % make
    

    The make command will automatically download, configure, compile and (locally) install external third-party source code useful to Conjecture, and will then proceed to configure and compile Conjecture itself. The end result should be a fully prepared environment, ready for OCR work. Note however that so far Conjecture has only been tested under Cygwin, Redhat 9, and Fedora 4, and may currently assume the existence of infrastructure not available on your platform - please inform us of any errors during this process.

    The only two external code-bases currently automatically installed are gocr and ocrad. Support for gocr has been more thoroughly tested than ocrad, and errors in compiling ocrad are not fatal to Conjecture itself (but errors in compiling gocr currently are fatal, but this is a temporary thing).

  4. Run the test harness verification suite:
    % make verify
    

    This will invoke the Conjecture test harness (consisting of multiple variants of multiple OCR modules over multiple images). Part of its output should be a table similar to the following:

        ...
        ============================================
        Name      4x6   5x7   5x8 ocr-a ocr-b   rod 
        --------------------------------------------
        default                                     
          v1        +     +     +     +     +     + 
        gocr                                        
          v1        +     +     +     +     +     + 
        ocrad                                       
          v1        +     +     +     +     +     + 
          v2        +     +     +     +     +     + 
        wmh                                         
          v1        +     +     +     +     +     + 
        --------------------------------------------
        Test Harness: Assessment of ./ocrprog
        ============================================
        ...
        

    The plus signs ('+') indicate the expected output was obtained. Any other symbol indicates some kind of problem. For example, of ocrad failed to compile on your architecture, the rows for ocrad in the table above will containg '@' symbols instead of '+' symbols.

NOTE: <somedir> is an arbitrary directory of your choice on your computer.

NOTE: An autoconf compilation environment will be provided in a future release.

Downloading from the SVN Repository

You can download the latest Conjecture source code from the SVN repository:

  • Verify that you have svn installed.
  • Read-only access is available here, thanks to SourceForge.net. To download it, use: svn co https://conjecture.svn.sourceforge.net/svnroot/conjecture conjecture. Read this for more information.
  • Send a request to conjecture-devel@lists.corollarium.com for write access to the repository. Please specify what username you would like (this username will show in SVN log messages, etc.) This username must be registered in SourceForge.net.
  • Wait for one of the admins to grant you write access.

Downloading Binaries

TODO

Quick Links

  Downloads : V-0.06
Repository
  Howto : Install
Modules
Implementations
  Community : Mailing List
Wiki
SVN
  To Do : Questions
Easy
Design Implementation Infrastructure

Conjecture is using services provided by SourceForge