Tensorflow 2.0

Calling out to @hunterjm as I believe (from the release notes on 0.82) he was behind the implementation of the Tensorflow component.

Are there any plans to update Tensorflow component to the new Tensorflow 2.0?
I’ve read on it a little, and they make it sound like it is easier to use (and so updating might be easier than the original development). But I have no confidence in my own ability to do something like this and so just wanted to ask.

I use the component now, and it is fantastic and so easy to setup!
My only issue is that I have a tree it is positive is a person at night :rofl:
Not complaining though, it is just funny to see.


TensorFlow 2.0’s changes are mostly in how models are created. Not much difference in inferences. Will probably update at some point, but it is still alpha.

Thanks for the reply!
So no performance or similar improvement?
Makes sense then to not be a priority!

Quick question - I have been working to setup my camera’s for motion detection - I know there is the rectangular “area” for some setups, but is it possible to use a mask to block our regions?

This would be for motion detection (FFMpeg, etc.) as well as object detection (to get rip of the above mentioned tree person :slight_smile: ).


Polygonal masks would be too difficult in HASS. For TensorFlow you can define an area and if the tree doesn’t fall completely inside the defined area, it won’t register.