Hexapod Makeover

[manga valk]

do you build them fore your work ???
so yes what kind of work do you do (i want such a job to )

[blegas78]

Intel took a professional video of me talking about the 3D printer used for building hexapods:

http://edc.intel.com/Video-Player.aspx?id=3096

[innerbreed]

great video. presented well. the 3D printing is something iv always wanted to do. i once saw a guy print two ball bearing rings looped together,
http://www.youtube.com/watch?v=gnB8vFc123Y
so it justifies just how complex these parts can be.


again great work with your hexapod. and congratulation with your work.

[Matt Denton]

Very nice, always good to put a face to a name

[blegas78]

great video. presented well. the 3D printing is something iv always wanted to do. i once saw a guy print two ball bearing rings looped together,
http://www.youtube.com/watch?v=gnB8vFc123Y
so it justifies just how complex these parts can be.

That printer I think is the next model up from the one I use. The one I have access to only used breakaway support material, but the one in the video uses a soluble support material. It sure would be nice to use that one though, as it would have saved me many hours of work, and many finger injuries

[blegas78]

I began working on face detection for my hexapod a week ago, as obviously inspired by Matt's iC hexapod

There was a SWE conference in Phoenix yesterday that I was preparing for, and Intel had a booth for their career fair, so they wanted a hexapod. I figured I would bring two, one for people to control via the sixaxis controller, and one with face tracking. I worked myself until 4 am yesterday morning completing the head and getting multi-threaded coding down for face-tracking and kinematics running simultaneously.

Although it would have been great to get the 3DOF head going, I really need to implement bearings as the plastic-plastic friction is too great to be really useable. As a result, I designed a quick 2DOF temporary head just to get things going with bearings I salvaged from burned servo motors in the lab.

I still need to modify some of the mechanical bits as there is a fair amount of slop in the mechanics, but it worked well enough for the demo. Here is a cellphone video of the week's worth of work:

http://www.twitvid.com/5BF27

When I'm completely happy with the head mechanism and with the code, I'll post a vid on youtube, obviously crediting iC, and I'll link it with Matt's work.

[Matt Denton]

Looks good Matt!

What face tracking library are you using? and what kind of frame rate can you achieve on your platform?

[blegas78]

I'm glad you like it!

I am using the HaarDetectObjects function in OpenCV. I didn't have much time to optimize the face detection part, but right now it is running at 5.5 fps. How fast is iC face detection?

[Matt Denton]

On a good day 25fps, however, its usually sits around 15 to 20

[blegas78]

Oh wow, that's really fast, certainly something I strive for though. Are you using the HaarDetectObjects function? I played around with decreasing the image resolution which speeds up the detection, but if I decrease it too much it it can barely detect any faces. I still have to play around with some of the code, but I hope to get at least 10fps.

What processor do you use at what speed? The 1.6GHz Atom is perfect for kinematics (up to 400Hz update rate), but not completely ideal for face detection.