I know there is already a feature request for Kinect V2 integration, that would be very useful. I'm just doing some testing with it for the first time now. It works well with NI Mate, much quicker pick-up of the skeleton.

First thing to note is that it gives more joint position than the V1. Could you add a kinect V2 node with these extra OSC addresses please. (I know I could retrieve them manually but always good to have a built in node).

As you may know the NImate developer changed his licence/subscription recently and I asked him what plans he has for future support/development. Not much other than continued support for now it seems, the GL deprecation will probably effect some aspect of the programme in the long run too.

So how possible would adding the skeletal recognition into Vuo be? It would give us something to rely on into the future. (I read that you are considdering using Open NI in a thread somewhere).


Notes from Team Vuo

Vuo Pro: 

No — available with both Vuo CE and Vuo Pro


●○○○ — Up to a few days of work


●●○ — Could expand community


Joe, I've converted your

jstrecker's picture
Submitted by
Feature status:
Waiting for review by Team Vuo
Open for community voting

@Scratchpole, I've converted your question to a feature request.

Related: Node set for skeletal tracking with Kinect.

We did a cursory test of NI mate 2.14's "Kinect for Windows 2" sensor. In "Basic" mode (with hands disabled), it outputs the attached 27 points. Not sure why it's excluding Chest and Pelvis and including some fingers, but that would be something we'd need to look at more closely when implementing this. Unless you happen to know.

Thanks Jaymie.

Scratchpole's picture
Submitted by

Thanks Jaymie. I'll borrow the V2 again and try to answer your question about random and missing points this week sometime.

Although NImate is good at what it does it's just a little bit flakey, randomly shuffling it's Syphon outputs is one such annoyance. I'm guessing what you see in your OSC monitor is another example of the flakiness. It would be so much better to have the skeleton tracking done within Vuo.

If I understand correctly from your other posts on the Skeleton Tracking feature request, you will likely not be implementing skeleton tracking from the Kinect, but from regular video cameras using one of the new ML libraries. That'll be great, other cameras are so much more flexible. If it's at least as reliable and accurate as what is currently available it will be a boon! I guess the ML is only going to get better as it is used more and more.

I received this response from

Scratchpole's picture
Submitted by

I received this response from Jesse at NImate: 'With the hands turned off they aren't supposed to be outputted. It's possible it's a bug. You can force disabling them as follows: Enable the hands Clear all the fields for the finger joints The fields should no longer be outputted Chest and Pelvis are specific to Kinect for Windows v2 Microsoft SDK, I think. Other sensors as well as libfreenect2 won't output those joints.'

So sadly this FR is redundant, NImate with Kinect v2 running on a mac does not output the extra joints as expected in the first image in this thread. You only get the same points as with the V1. NImate on Windows should output the extra points.


A little follow up info from

Scratchpole's picture
Submitted by

A little follow up info from Jesse. Unfortunately Microsoft does not provide a Mac driver for their SDK and only libfreenect2 can be used on Macs. On a mac you are unfortunately limited to the Kinect 1 joints. If you have access to a Windows PC you could try using NI mate on it.


ranjithshegde's picture
Submitted by

@joe I am really interested in your trial. Although this was 1.5 years ago. I am trying to run NiMate (1, on linux arch) with kinectV2 and it does not detect device (libfreenect2 installed).

Did you make any changes to NiMate or its the last available linux(deb) version?

Feature status

  • Submitted to vuo.org
  • Withdrawn by reporter

When we (Team Vuo) plan each release, we try to implement as many of the community's top-voted feature requests as we have time for.

Read more about how Vuo feature requests work.