Personal ML Model on Home Assistant OS

Hi everyone!
I studied data science and I would like to introduce some deep learning network to my surveillance cameras.
Is there any way to introduce and let communicate tensorflow or pytorch models to Home Assistant os?
I read about foods and deepstack but, as far as I read, it works with docker only.

HAOS is most a command and control software.
There is not much altering or analyzing going on in HAOS.
The way other does it, like Frigate, is to install it as an addon in a docker container or as a separate installation on some other system.
HA can then command and control that docker/installation.

So, there’s no way at all to realize some artificial intelligence on the camera images on HAOS?

Not directly on HAOS, but in an addon there is.
The Frigate link I gave you has a link to the HAOS addon of frigate. It is a bit down on the page.

You have to remember that HAOS just shows the picture/stream from your cameras.
It can do that directly from the cameras or from a NVR/proxy, like frigate or whatever you come up with.
HAOS might even be able to control that NVR/proxy.
This setup means you can run the NVR/proxy on a more powerful machine and utilizing Google Coral USB sticks, Nvidia and AMD graphic cards and whatever might be of interest for you.

1 Like

I was able to successfully run Frigate! Thank you!