Hey everyone!
Got some free time, so wanted to think out loud about something.
I was recently looking into what's new in the microcontroller space. ST has been firing out new series like crazy lately. Stumbled upon the STM32N6, a series with neural network and camera support. What's interesting is that the chip clearly has enough resources to support Matter and a Camera Device, and on top of that do object recognition on video thanks to the built-in NPU.
But that's all just background.
What about using a camera in a slightly unconventional way? Not just as a device that streams video, but as a smarter motion sensor with a brain. React only to specific types of objects, people, animals, cars and only when they appear in a defined zone. Basically what expensive security cameras already do, but as a native Matter device, no proprietary clouds involved.
And that leads to another question, how interesting would a bridge be for existing cameras? Something that takes a stream from a regular IP camera, analyzes the video locally, exposes the camera in Matter as a Camera Device, and at the same time publishes a separate Matter Occupancy/Motion Sensor with smart filtering by object class.
Essentially a Matter bridge for legacy cameras. No ripping out hardware, no cloud.
What do you think, is there demand for something like this? Or is it easier to just replace the cameras with native Matter devices?