Jesper

Definitely! Seems easy enough to implement. Do you know if there are any pre trained models or do you have to train your own?

@sdwilsh Is this still of interest for you?

Looks good! One issue tho, i wont be able to build and push this locally since i dont have an RPi4 to do this... Not sure how to get around...

Sorry this is taking so long for me to implement, i cant find the time right now to finish this. Thanks for maintaining your fork in the meantime

@alceasan I finally got my multiarch builds to work, i put up a test-version under ```roflcoopter/viseron-new:docker_revamp``` Could you try and see if it works on your RPi4? Currently the generated...

> ImportError: libatomic.so.1: cannot open shared object file: No such file or directory Fixed and pushed a new version of the same tag, Could you try again?

> Sorry about the delay, still not working, this time a problem with EdgeTPU libs: > Dont worry about it! I am the one whos late... ;) Could you exec...

Actually scratch my last comment, it wont solve anything. The error printed is just the caught stacktrace when no EdgeTPU is found. Did you start the container with ```-v /dev/bus/usb:/dev/bus/usb...

> The container is already running with that configuration, but I'm not using any Coral USB. Aha i see! Well that means it should work then, because the stacktrace you...

The container is built with `v4l2` codec, as you can see [here](https://github.com/roflcoopter/viseron/blob/99ad9ae4ec7f7838ecf0d24240179549a0a9f455/docker/aarch64/Dockerfile.ffmpeg#L497). However, i found a typo in one of the environment variables so the codec was not automatically picked...