ZED camera body tracking alpha mask

We’ve been making great use of the zedcam plugin in the latest 2022 release. The tracking is fantastic. However it would be amazing to also get the mask for each body tracked. This is available as an output from the SDK, and is one of the few things preventing us from completely switching over from the kinect:

Is there any plan to expose this as a TOP operator? e.g. a single texture, with the pixel values set to 1 for p1, 2 for p2, etc?

1 Like

Hi @AirbagProductions,

thank you for the suggestion. It’s been logged as a RFE for consideration.



i would also realy like a feature like this for the zed 2i, we are trying to make our own mask from the zed sdk and some python in touchdesigner but it is proving more difficult then expected.
a “player index” function like the kinect where each player has thier own color would be super handy

1 Like

The way the ZED SDK is setup, they provide a different texture for each detected object. Do you see more value in that workflow (with a way of selecting that texture), or a singular texture we’ve I’ve already merged them all together for you?

1 Like

Hey Malcolm, that’s a good question, thanks for asking. My (slight) preference would be for a single texture with all human body masks merged together, but only if identifying information was embedded in the mask pixel values - e.g. the skeleton ID/player number, so that masks can be matched to body tracking data using our own code.

But overall, as long as there is a way to associate body tracking chop data with mask data, then we’re happy on our end.

1 Like

Hi @malcolm , just checking in to see whether this RFE is likely to make it into a release any time soon? We’ve got a few projects this could be helpful on - I’ve been playing around with google’s mediapipe as an interim solution, but it’s a bit hacky. Thanks for your help.

We actually have it 95% done, but tracking down a crash that is occurring in some cases still.

1 Like

We’ve confirmed with ZED that the crash is in their SDK, so we need to wait for SDK 4.1 to be released, then we can release this feature