Which Kinect model and operating system for easiest use of Kinect with openFrameworks?

Dear community,

unfortunately I didn’t find a definitive answer which is not older than almost one year … it seems Kinect v2 is still not officially supported by ofxKinect, is this correct?

So I just wanted to ask you experts, as I never used a Kinect by now, which Kinect model I should get and if it’s totally fine to use it with macOS, or if I should use it with bootcamp windows (10?) or even with a real windows laptop …

I’d like to use the Kinect to measure the speed of a body’s motion and maybe the distance to the camera (for now) …

Thank you very much!

PS: I’m also not sure if I need a power supply for Kinect (for xbox 360) when using it with my MacBook Pro …

Hi, you can use ofxKinect for any Kinect V1 Kinect.
For Kinect V2 on maxOS use ofxKinect2 ( https://github.com/ofTheo/ofxKinectV2/tree/feature-worldcoord ) or https://github.com/hanasaan/ofxMultiKinectV2

You will need a Mac with a USB 3.0 port and you usually have to turn the power off and back on on the Kinect after it is plugged in for the first time.

Dear @theo, thank you very much for your answer!

I already stumbled upon your ofxKinectV2 but wasn’t sure, if it’s still up to date because of the latest commit being two years ago …

Are there any advantages/disadvantages in using Kinect 1 or Kinect 2? In general and, for example, in terms of stability and liability.

Thank you so much!

Kinect 2 performs much better. There seems to be a general agreement over this… if you can choose go for V2

For longer distance depth data and for running multiple kinects reliably and for running long cable lengths I would recommend V1

Kinect V2 has finer depth resolution but doesn’t do as well at longer distances. Its color image is HD vs V1 640x480

So it really depends on what you want to do.
But V1 is generally a bit easier to work with.

1 Like

In case you go for V1, I would strongly recommend to get a Model 1414.
I really had a tough time getting a 1473(?) to run. 1414 ran out of the box on every system and OF version for me so far.

have a good day!

Dear @davidemania, @theo & @dasoe,

that’s some pretty solid advice so far! Thank you very much. So I’m pretty new to this camera/kinect topic, that’s why I’m asking that much … I came up with a project for my university graduation and it will be aligned somewhere between choreography and, well, technology. I want to use the kinect—maybe it’s even the wrong tool for this approach, which would be a shame (to my research)—to track and/or measure four things in the best case.

  1. Where the choreographer is “positioned” in the room … meaning from left to right. Think I’m talking about a distance of 7–10 meters here …

  2. If the choreographer is moving or standing still and kind of like the “speed” of the movement … meaning not exactly how fast she or he is moving, but metaphorically speaking on a scale of 0 (standing still) and 10 (moving very fast).

  3. Maybe tracking the position/movement of the choreographers hands and or feet … I think this could also be done with some markers if otherwise too complex.

  4. Maybe the distance between kinect and the choreographer.

Do you think this could be possible? I doubt the resolution could be enough for a distance of 7–10 meters …

Hi fooness,

(IMHO): The kinect will do 1,2 and 4 quite solid. You will get a 3D point in space where the performer is (1 & 4). And of course you can then monitor the changing in this point (2). (Note that there are multiple ways of achieving this. You can go with the software people recogintion. This is a great and easy thing: it finds the complete skeleton for you. BUT it might need some time or loose the person for some seconds in between. The other approach- in case you do not really need skeleton - would be to use the kinect as a depth sensor and do the rest yourself. For a live performance that should never fail I would at least think about the latter approach. But this is a different topic).
3 will also work, but might be a little less reliable: for this you’ll need to use the software people recogintion and also a hand is even more difficult to track than the body.
Problem I see is the distance, though. I would not suggest to use more than 4/5 Meters, depending on setting and lighting. Kinect will not see anything more far away. Maybe you could use 2 kinects instead but closer?

hope this helps!

Dear @dasoe,

thank you very much! That sounds quite good and gave me some optimism about the project … I could also use two kinects, or even three, if it’s better to go this way … I cannot yet quite imagine on how to work with two or more kinects at the same time, but that’s something for experimentation in the next three–four months I guess …