[GRASS-dev] Fwd: Image Segmentation - GRASS - GSoC report #5

[Eric Momsen]

Hello,

GSoC report is below.  Also when googling a grass function, I found my code
diff's in the list archives.  I hadn't made the connection that changes in
the extensions are also forwarded to everyone that keeps track of source
code changes.  So I hope my sometimes multiple commits in one day aren't
considered spam?  Let me know if I should hold back and only commit 1-2
times per week when more significant tasks are completed.

Have a good weekend,
Eric

---------- Forwarded message ----------
From: Eric Momsen <eric.momsen at gmail.com>
Date: Fri, Jun 22, 2012 at 4:21 PM
Subject: Image Segmentation - GRASS - GSoC report #5
To: soc at lists.osgeo.org


Hi all,

Completed:

Both planned tasks were completed:
* using linked lists to track segment membership
* using preexisting information (raster areas, original described as
polygons) to constrain the segmentation process

Also Markus M was running the module, and uncovered an error in the
algorithm, so next week's plan for validation is already started :)

Plan:

* Further validation to confirm the segmentation results.
* Find some way to improve the speed, with improvements to the current code
and/or move more data into RAM.
* Start a second module, to produce statistics for the segments.

(This might be too much...hopefully at least 2 of the 3 tasks can be
completed.)

Blocking:

Nothing really.  I have some new appreciation for the number of
calculations that must be made, the established closed source software runs
overnight to segment a typical image....even with that basline I have a
long way to go, right now I think the current version here would take 2
years... hurrah for challenges :D

Best regards,
Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20120622/0698a68d/attachment.html>