Design Flaw in AI detection

Actually the bug/design flaw is in ANY detection setting, but Vehicle is the most easily demonstrated. If there is a vehicle parked anywhere in the detection area, ANY motion will trigger the vehicle detection. The proper design is that only the item in motion should be evaluated for being a vehicle, person, pet, package, whatever.

The way it works is actually any motion will trigger a cloud event. Once the event is uploaded to the cloud, the ai then checks the video for recgonized objects, weither moving or stationary, and tags the event as such. The tags are not necessarily what the detected motion was, it’s just what’s recgonized somewhere in the video.

There are multiple wishlist topics for improved ai, here is one that may be one you want to support. Visit and vote and comment to help it along!


Wyze did a survey recently on what they preferred for detections. I think they said it goes until this Friday when they will draw winners, so you probably still have time to add your opinions to it. You can find the survey link in this thread:

Wyze will get your feedback on this preference there.


So you only want to know when a person, package, pet, or vehicle first appeared in the camera’s FOV? That seems like a solution that would lead to worse unintended consequences than the core problem… namely that there are too many motion events triggering by stuff you aren’t interested in. Not that you only want objects tagged in the recording in which they first appear. I don’t even know how Wyze could pull that off. And it would set off a firestorm of legitimate complaints like “I didn’t have any events or notifications for the entire 5 hours that these strangers were on my property!”