Flags being detected as packages and people

I get notifications on our front porch and yard cameras that a person or package is detected. But the only thing I see in the clip is the flags or furniture we have on the porch. It doesn’t happen all the time or I’d have to disable the functionality.

Oddly seat cushions sometimes cause a package detected. I had to turn the sensitivity down to reduce the false positives.

Detection zones don’t seem to work, and there is nothing to suggest flag or furniture detection except tree moving or light change.

Is there a way to train for recognizing flag activity?

The only way to train the AI is to submit the Video to Wyze and change the tag in the dialog.

However, it is very possible that some other object in the frame is what is being tagged by the AI Engine.

Wyze does not mark the video to identify what object in the FOV was identified and tagged. If they did, our lives would be much easier.

The only thing that gets marked in any video is the green Motion Tracking box and that is applied to the video by the cam prior to it ever being uploaded to the server AI Engine.

While the Flag may be the moving object that is triggering the Motion Activated Upload, it may not be the object that was identified as a package and tagged. That may be why setting a Detection Zone isn’t excluding the flag… because it isn’t the “package” object.

Lowering the sensitivity isn’t the best option because it severely restricts the effectiveness of the cam. The better option is to consider the FOV and place the cam to view an area that isn’t going to have constant motion from its surroundings. Next is to use the DZ to exclude high motion objects that would continually trigger Motion Activated Event Uploads.

Once this is done, a very close inspection of a snapshot of a Package Detected Event needs to be done to identify every object that might resemble a box. Then either remove the object from the FOV or use the DZ to block objects one at a time until you find which object is being habitually tagged.