Machine Learning - Teaching Isadora to recognise different human shapes

mark_m

Hello Isadormobiles,
I've been following @Andy-Carluccio 's tutorial on Teaching Isadora to recognise images, object, people, which is here:
https://support.troikatronix.c...

I've run into a couple of issues - perhaps not surprising, as Andy wrote this five years ago!
1. The tensorflow models produced by Teachable Machine are now in .js format, not the .h5 that Andy's NDI Classify program wants.
With help from Claude AI I did convert the Tensorflow model to .h5

2. When I run the NDI_classify.exe I briefly see a window flash open, and that's it. No dialogue box as expected. Tried on two different PCs, so...
@Andy-Carluccio in the instructions it says "run NDI-Classify.exe" I didthis just by double clicking. Is there a particular way I should be running it?

Time has moved on, and perhaps this is no longer the best way to achieve this? I have this well trained tensorflow model which I want to use in Isadora to trigger events when the different human poses are detected. Shape A triggers event A, Shape B triggers event B, etc.

Any thoughts on this?

Thanks all!!

Mark (not...)

DusX

@mark_m

I haven't gone through this in some time. I will go through this tomorrow, and see what needs to be updated to make the process smooth once again. My gut tells me it would be better to use the current format/s supported by Teachable Machine, and update the libs / code used to work with the current format.

mark_m

@dusx

Thank you. Managed to do this with the help of Claude Sonnet 4.6 which showed me how to retrain the data using something which has the same landmarks as in your Mediapipe pose tracking examples.
This seems to work very well, thank you.

Not sure how to share the process?

I took maybe 25 photos of each of 14 poses (yes, different handstand shapes, of course!)

Then organised them into this folder structure

training_data/
arms_crossed/
photo1.jpg
photo2.jpg
...
arms_raised/
photo1.jpg
...
lunge_left/
...

And then followed Claude's instructions.

Here's my Colab notebook:
https://colab.research.google....

After I'd successfully trained that and got the files out the other end...

It then gave me the python mediapipe code.
https://www.dropbox.com/scl/fi...

And that outputs the OSC into Isadora. Two streams, one telling me which pose it is, and the other giving me the confidence about it.

Hope this is useful to someone!!!

Cheers

Mark (not that Mark, who I'm sure doesn't use Claude to do his coding!)