|
[DPRG] EmbedCV version 0.2 released
Subject: [DPRG] EmbedCV version 0.2 released
From: Chris Jang
cjang at ix.netcom.com
Date: Tue Nov 28 20:16:22 CST 2006
Hello,
The computer vision library I'm developing is checkpointed at version 0.2.
There is a website now.
http://golem5.org/embedcv/
There are detailed demonstration pages of image segmentation and object
detection. There is also a high level tutorial of some basic concepts in
computer vision. I tried to avoid invoking mathematics so hopefully it
makes sense.
The code is almost a total rewrite from the last version. The library is
around 5000 lines of pointer arithmetic and bit shifting. There is another
1500 lines or so of sample command line tools. They try to emulate the
utility of the ImageMagick tools except for computer vision specific
algorithms instead of general image processing.
A lot of functionality has been added. The stuff I'm proud of are the fast
morphological operations and the object detection. Object detection always
used to seem like magic. It's really not. If you look at
http://golem5.org/embedcv/example_obj.html
you will see this. Next comes object recognition. But that will take much
more work.
Feedback from several individuals has convinced me that I need to take the
plunge and have a GNU autoconf configure script. There are also a few
places where compiler warnings may occur. All of these enhancements will
go in the next release. I'm figuring that will happen in one to two
months.
I have to try integrating the library code onto my robot. If this works
and it can follow people, etc, then that will validate the embeddability
objective of the library. EmbedCV is still very immature. But I feel there
is enough at a sufficient level to be useful for robot applications.
As always, feedback is appreciated.
Chris
More information about the DPRG mailing list
|
