Implement basic architecture for image processing with ME / TF

[pa-wills]

MotionEye is giving me URLs for streaming and for still. Use them.

Configure a MJPEG IP Camera entity from the URLs: https://www.home-assistant.io/integrations/mjpeg/

That alone should get me the basic images into HA.

Then - configure Motion to send call a web hook when motion is detected. Less sure about this, but it will indeed work.

The implied Web Hook handler on the HA side would then need parse the output and update a state variable. On that - is integration to nabucasa required? Out and back seems a bit silly, but the alternative is some sort of local API, I suppose. There’s probably a way to expose a web hook locally from HA, and leave it at that.

OK, now I have a detected event. From here it should be fine to pick up the trigger, go get the latest jpeg and start running things through TF.

A different route - do the same type of thing with ZoneMinder, perhaps. I would rather be on ZM, but is that a true replacement for ME? Not sure. OK, yes, it probably is - but we’re not going to be able to start here without a NuC, and some exterior cams. Worth acknowledging the target state though:

• NuC (for ZM and for TF)
• Lots of IP cameras. Power-over-ethernet hub.
• Some sort of integration from ZM into TF into HA. Possibly skipping TF in the event that the zmMagik and EventServer integrations supplant it.
• Capability to do object detection (are my bins here, or here?), (where’s my car), etc. converted into events. From there you’ve got the triggers you need for automation. All of this is about creating triggers, ultimately.

[pa-wills]

I am beginning to think I should just try to get this done initially in the cloud. Say with AWS Reckognition, etc. THat would serve two goals - advance the project and advance my learning of AWS, which I have decided to invest in (development time).