The “normal” way that I’ve been doing occupancy detection is using motion sensors - if no motion has been detected for X amount of time, assume no occupancy, otherwise there is someone there. This works fairly well - until I curl up on the couch with a good book, or to watch TV or the like. At that point, no motion is detected, and no occupancy is assumed (incorrectly, of course!).
It occurred to me that Since EzVidoo has object - and person - detection capabilities, perhaps a camera could be used for occupancy detection? The idea would be that the image would be checked for objects - specifically people - every so often (say, every 30 seconds or 1 minute, perhaps user configurable), and if a person is detected, then assume occupancy, otherwise if no person is detected in multiple checks over X minutes, assume no occupancy.
In essence, there would be a “person” sensor associated with the camera that behaves exactly like a motion sensor, except it is triggered by the detection or absence of a person in the image. This sensor could then be used in meshbots in the same way as a motion sensor would.
I actually implemented this concept externally using Python and PyCoral (https://coral.ai) on my driveway security camera to excellent effect. I have that running at 5 frames per second so I get immediate notification of anything interesting in my driveway - though there I look for more than just people (cars, trucks, cats, dogs, moose…). It would be really cool if it could be integrated with the Ezlo ecosystem, however, so I could use it to do things like turn on lights if it is night.