Status of computer vision in SOAR?

[inf...@fwsudia.com]

In the recent book, The Soar Cognitive Architecture, Laird said that integration with some vision system, such as Open CV, was in the works. How is that coming along? A search for the word "vision" through most of this website and wiki turns up nothing. Thanks

[Mazin]

We have a system we developed called SVS (Soar Vision System).  It might make it into the next version later this spring, but we're not sure if it'll be release ready by then.  You can contact Joseph Xu, who wrote it, for more information.  I believe there is also some information about the module in John Laird's recent book on Soar.

- Mazin

On Fri, Mar 15, 2013 at 6:40 PM, <inf...@fwsudia.com> wrote:

In the recent book, The Soar Cognitive Architecture, Laird said that integration with some vision system, such as Open CV, was in the works. How is that coming along? A search for the word "vision" through most of this website and wiki turns up nothing. Thanks

--
You received this message because you are subscribed to the Google Groups "soar-issues" group.
To unsubscribe from this group and stop receiving emails from it, send an email to soar-issues...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Frank Sudia 974]

Mazin,

With some further clicking I found Xu's slides in the last Soar conference, and his SVS source download, however it was not linked anywhere on your main pages. This is a key feature I've been waiting for, and I like Xu's overall approach. It was also mentioned in the book, which was how I knew of it.

Thanks,
Frank

You received this message because you are subscribed to a topic in the Google Groups "soar-issues" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/soar-issues/NmYsgVyPYKI/unsubscribe?hl=en-US.
To unsubscribe from this group and all its topics, send an email to soar-issues...@googlegroups.com.

[Mazin]

No problem.  FYI, I think it's been developed quite a bit since last year.  I'm not sure if the code under that link is updated or not.

[Frank Sudia 974]

I didn't appear to be updated, he called it SVS-master, but not to worry, as I'm just looking at it, not attempting to compile or use it. I'll most likely wait for the release, whenever that is.

[Joseph Xu]

Hi Frank:

On Sat, Mar 16, 2013 at 3:17 PM, Frank Sudia 974 <inf...@fwsudia.com> wrote:
> I didn't appear to be updated, he called it SVS-master, but not to worry, as
> I'm just looking at it, not attempting to compile or use it. I'll most
> likely wait for the release, whenever that is.

Sorry, I've been out of town for the last couple of days. That branch
hasn't been updated in a while because another project was using it.
I've been secretly committing to a branch local to my computer for
quite a while now, but those changes will be pushed into master before
the release. Can I ask you what you intend to use SVS for?

Also, let me clarify a few things, in case there is any
misunderstanding due to lack of documentation:

* SVS doesn't use OpenCV and doesn't provide similar functionality as
OpenCV. The V in SVS stands for Visual and not Vision, because SVS
doesn't do computer vision stuff like object recognition. The input
SVS expects is geometric information (such as the positions of objects
in 3D space and their shapes as convex hulls), not images.

* Samuel Wintermute is responsible for the design of most of the
functionality of SVS. I only reimplemented the system to make it
easier to use and compile.

Joseph

[Frank Sudia 974]

>

Can I ask you what you intend to use SVS for?

I've been considering ways to reason about objects in an actual scene, like of some physical blocks on a chess board, which would require OpenCV type functionality. Also I believe the recent book by Laird stated that some type of vision capabilities were in the works, that would draw on external projects such as OpenCV.

However, it looks like you're starting with internal "visualization," in the mind's eye, like Winograd's famous Shrdlu in 1971, which also implemented a 3D blocks world model, now over 40 years ago. Maybe check out the later Java port of that system for ideas.

Can't any Soar-bots see? Is this capability even on the drawing boards?