VisionState - New App for creating sensors from camera feed. (Example Garage door closed)

I started creating an App for home assistant with the goal of it being easy to create a sensor from a camera feed, my use case was my garage door position.

Ive tried other solution but had a hard time getting them to work.

VisionState installs as an app on home assistant, runs locally and on the CPU, so no dedicated GPU/NPU/TPU required. It integrates with home assistant with MQTT and can use home assistant devices as triggers.

Its early in development, but feel free to check it out and come with suggestions/improvements:

GitHub - oleost/VisionState: Teach a local AI what your camera sees and turn it into Home Assistant sensors — garage door open/closed, gate shut, car parked. 100% local, any CPU. · GitHub**

whouha, great idea, let me test it !

Would love there hear how it went

I’m pleasantly surprised so far; the detection is very fast, too.
My use case isn’t simple:
I use it to detect the presence of our two parked cars. They are quite different from each other and can be parked in different ways.
The training is working well for the moment, though I still have a few “parking” scenarios to teach it before I can say everything is working perfectly!

Happy to read, nice :ok_hand:

the newest update supports face recognition.