Florian Echtler

(and could also be used for calibration).

For the (beta) reference implementation, see here: https://github.com/mkalten/TUIO20_CPP It's still based on OSC and also has native WebSocket support by now. Maybe our upcoming project can work on this, what...

Probably related, I just noticed that a code comment on a merge commit (i.e. not directly associated with the PR) also does not seem to be backed up.

Interesting question - this won't work out-of-the-box, because all the ML models are trained to segment _persons_ and not _faces_. But you could try a two-step approach, where you take...

Ah wait, now I remember that body-pix actually does separate classes for body parts, including head. That might work _almost_ out of the box.

Hmmm, if I remember correctly, @BenBE had some code that would just run the model twice for wider images, but I'm not sure if that ever made it to the...

You would need to use `libcamera` directly in reacTIVision (cf. https://github.com/libcamera-org/libcamera/blob/master/Documentation/guides/introduction.rst). Not sure right now where in the code this should go, though, it's been a long time since I...

P.S. apparently there is also a wrapper script `libcamerify` which already does something similar, you could try launching reacTIVision through that.

Alternatively, pick the most common color in HSV space: https://docs.opencv.org/3.4/d6/dc7/group__imgproc__hist.html#ga6ca1876785483836f72a77ced8ea759a

Hm, I think the only way to solve this would be to run SmcDumpKey on macOS, once with TDM active and once with TDM off, and compare the results... or...