pouët.net

hey guys,

just wanted to show our 'demoish interactive kinect installation':

https://vimeo.com/51247057

we also have plans to bring it to some demoparties.

That is so cool! More about the setup and the tech? I'd love to do something similar.

Beautiful. Tell us more!

Awesome! Bring it to Evoke!

Cool :)

there are 4 xbox kinect cameras positioned at the corners of a rectangle in front of the bigscreen.
the biggest problem was to combine the data of all kinects. it's now pretty stable, in that setup 2 cameras are attached to the ceiling, 2 are hidden next to the bigscreen. so the view angles are very different.
to adjust procedure requires a special "adjustment dance" so the software can match the different point clouds and combine them.
you can now place a virtual camera behind the "players", in this setup it is just used as a mirror basically.

the combined data is broadcasted over network, so different sound/graphic engines can read the data and display graphics/create sounds etc.
we also realized a project that streams data over the internet, so dancers in germany and china can dance together/communicate visvually.

we also have some more effect, which use the skeleton features (at platine only point cloud data). you can see some development screenshots here: http://tom.undev.de/

we plan to show this at some demoparty, not sure when/where...

What backend did you use? The Microsoft one?

we use the microsoft sdk, but that can only get data from one kinect.
everything else is our custom code.

looking good

The remote streaming sounds awesome... especially if was >2 sites with rendering done in different colours for each site...

nice, sent it out to my students :)

So wait what, you had 4 PC's running for this??

Very cool!

So it requires an adjustment dance does it? Is it anything like this? : https://www.youtube.com/watch?v=6pOq4hyoX9g ... If so I guess it's about time to remake that old classic with a few subtle changes to the lyrics :).

is it just me who's actually somewhat unimpressed? i mean from purely from a visual perspective it just looks like "fill shape with random polygons and dots and lines"

it's the sync with sound that makes it interesting

<3 <3 <3

Superb!

I hope you will be able to bring it at Revision.

Gargaj: the "shape" that's being filled is a 3d captured model of a body, which can be also be seen by the not-so-randomly-placed polygons. At least that's how I perceive it. It wouldn't look half as good if you'd only have a 2d filled silhouette of your body. So the real challenge here was indeed the capturing.

Visualwise I tend to agree, that it is not impressive at all, but the overall product is.

xtr1m: generating a model in 3d (..ok, 2.5d) from kinect data is pretty straightforward.

finall kinect is put to some good use.

From a technical point of view I'm with Gargaj and smash.
Art-wise this is interesting.

But I guess one has to "experience" that thing in action in order to properly judge it - which is a bit difficult by watching a youtube video. :)

Smooth real-time surface reconstruction from the noised sensor data is a bit harder.

...and consolidating the data of four kinects into a single playpen might also be trivial in that regard, but did you see those two fragments of a kid wildly prancing in front of the screen in pure glee? Magic. Fuck your Vulcan logic.

I'm just filled with questions regarding the screen and the angle of the kinects and projectors.

Shifter: yea, the exact same reaction I've seen at a bunch of interactive installations... Wooooo, im on tv messing with the pretty picture.
The problem with that scene is its still so shallow. I want to see it move past little toys that look pretty towards something that goes deeper, makes you think and is worthy of the "art" tag they are stamping on everything.

smash is jealous. he ruined cncd name with spacecut and people started hating him for that. or it was ercola. real talk.