Wyze stated that they have sold over 3M cams now, which does require some serious number crunching, that I agree, but, highly doubt it has to do with “overloading” a server, this is virtual compute, and they can (and have to) allocate resources on the fly.
I am guessing that they are seeing the minimum processing time required, so they don’t buy more than they need, but, problem here is, they won’t get an accurate picture of usage if they artificially cap the testers.
They should be adding hundreds of people a day to get a more accurate representation of usage patterns, or this is all going to be one huge problem on launch day when you do get those 3M cams hitting the cloud all at once.
Mixed success with person detection. I have 1x Pan, 1x V1 and 2x V2 cameras. All have been updated to the latest firmware, motion tagging removed, person detection turned off and on for each camera and waited an hour. Here are the results:
Pan & V1 - no person detection at all. I thought I had read somewhere that person detection was cloud based and wasn’t limited by hardware so should work for these two cameras?
V2 Models - both are working and providing person detection with good success
I’ve tried restarting the devices, adjusting detection settings and all sorts to see if that would help, but no change.
People are detected when they walk up my drioveway toward the camera.
The same person is not detected by my front door camera. Maybe, because of the camera angle, they aren’t walking in a direction the camera would perceive as a person.
I think it’s more of a distance thing. Notice in the first video the person is fairly close before the Person Detection kicks in. The bottom video the person was on the far edge of the view.
Yup, could be.
The tracking rectangle doesn’t show until the person is about 10’ from the front of my garage (Video 1).
The distance from my front door to the person on the driveway is about 25’ (Video 2).
I honestly don’t know for sure, just a guess. But I notice I get motion sensor videos well before person detection videos on my back deck camera which gets a lot of foot traffic. Mostly dog walkers. The person detection videos seem to kick in when people are about 15 feet from the camera. The motion sensor videos seem to kick in at about 30 feet.
I think person detection is gotten really well. I have three cameras in every time I walk through a room or my wife walk-through room I get an alert. So I think it’s done very good
My person detection seems to be working pretty well. The only comment I have is that I’m getting a fair number of false alerts. The usually parallel a separate motion and seem to occur mostly at night, although that last is more of an observation than a quantified statement.
I have it like that so I can see the door, step and walkway.
The field of view is wider than it is tall.
Are you assuming that the orientation of the camera is the problem or has Wyze told you that?
Either way, the camera should detect people in any direction.
That is the problem, the models aren’t setup for just ANY orientation, just the standard one. The same issue with Xnor.ai, detection rate is very poor for the way you have it setup.
If you throw tons of processing power, sure, it can then detect it in any orientation, but, that costs way more than the standard method.
Well then I guess that camera won’t be detecting people other than Labrador Retrievers. Perhaps I can get the delivery people to get down on all fours.
It did recognise my wife going out the door last saturday, older firmware, but didn’t recognise me walking up to the door this morning.
