howto.standalone

Please read your package and describe it at least 40 bytes. System will automatically delete the di

How to use Dexter outside of the ADS
====================================

Dexter was primarily written to support extraction of data from scans
at ADS.  However, with a little fiddling you can make it work on
your local site.  Rather than doing this -- in particular if you've
never built a Java application before --, I suggest you check out
http://vo.uni-hd.de/dexter/ui/ui/custom.  This is a version of dexter
you can upload your own data to.

The instructions below only cover Unix-like systems.  
You should make sure that you have fairly recent versions of
Sun jdk (version 1.5 or higher), GNU make, and that both javac 
and java are on your path.  

As far as windows is concerned, you may be able to run Debuxter if you
have a local version of the JRE installed, but the rest of the
scripts in this distribution will be useless to you (some may 
actually work if you have ghostscript and cygwin installed).
What should work under windows is opening a command line prompt and
then typing:
     
     start Debuxter example.gif


Debuxter
========

With debuxter, you can extract data from bitmap images of (preferably)
pre-cropped graphs.  Note: Indexed images (GIFs, some PNGs) should
not have more than 128 colours, since otherwise the recognizers
won't work due to a bug (or my stupidity) in AWT's IndexedColorModel.
For the usual graphs, 128 colours are more than enough, though.

Try

make debug

in the directory you unpacked the Dexter sources to.  When the machine
is done building, you should see Dexter's extraction window.

If that doesn't happen, your Java Runtime probably is too old.  If
you send me bug reports, please include the entire output of "make
debug".

To process own images (but see above),
java -mx30000000 -jar Debuxter.jar <name-of-image>

(the -mx option just sets the heap size of the program to something
reasonable for handling images).  You should now see Dexter's main
extraction window with your image in it.


On windows, you'd put both Debuxter.jar and your image into a common
directory, open the command prompt, go into that directory and say

start Debuxter.jar <name-of-image>

You may have to increase the heap size here as well.

Note again that Debuxter doesn't support saving the results.  You can,
however, "Show data" and cut'n'paste the results.  See the following
paragraph for a somewhat more streamlined stand-alone operation.


Gaucho
======

Gaucho gets its images from a shell script, and thus you can potentially
extract from anything you can get images from.  Due to the dependence
on shell scripts, this won't work on Windows unless you get yourself
something like Cygwin (which is quite recommendable anyway).

Probably the most common application is extraction from postscript files.
The script dexter_getImage that is shipped with the distribution uses
ghostscript and netpbm for this conversion -- you need to have both
packages installed for Gaucho to work.

Go to the directory you unpacked Dexter to and say

make test

After a while (possibly quite a long while) you should see a selection
window pop up.  Mark the part you want to run the extraction on with
the mouse, wait a little more[1], and you'll see the selection window.

If this doesn't work, it may be that you're missing tools -- check
if there's a convert.log file and check its contents.  Also, try

dexter_getImage -s8 sample.ps

-- this may help to see why the image aquistion failed.

To extract from your own postscript or pdf, say

./runGaucho /path/to/your/ps.

This extracts from the first page.  If you want to extract from later
pages, just prepend a <pageno>@, e.g.

./runGaucho 4@/path/to/your/ps

extracts from the fourth page.

If you know a bit Bourne shell, you can check out dexter_getImage and
fiddle around.



System-wide installation
========================

I've made no provisions for system-wide installation yet, but if you
ask me, I'll probably have motivation to do so.

---

[1] This can actually take quite a while if your machine is slowish
and your extraction resolution is high (you can probably speed it up
up a bit by reducing the AlphaBits in dexter_getImage:procVector 
at the expense of antialiasing).  This could be made much faster
if ghostscript could be told to output a cropped image.  Any hints
in that direction are welcome.