Does Person Detection Override Detection Zone?

On one of my Pans I have a small Detection Zone set, Detection Zone On, & Person Detection on (via the Wyze Service page).

Today is a windy day & I am getting lots of Motion notifications, but the green rectangles highlighting the area of movement are outside the Detection Zone I have set. It appears to me that the Detection Zone is being ignored.

I believe I’m understanding you correctly, the motion outside the detection zone should not be triggering recording. the detection zone is what triggers the camera to begin recording. once it has begun recording, the motion detection ( green rectangles) will move anywhere within the frame of the camera. so once you have the recording detected, if say a person or the motion were outside the original detection zone, the rectangle would still follow them so you know precisely where the motion is. it should not be triggering the recording unless the motion is within the detection zone at some point. and being shadows that could be just a flash of movement that triggers the pixel change.

I would look back on your videos and see if any any point if the green box originated in your set detection zone, even for a second, enough to change the pixels be that by shadow or other movement and thus leading to the recording.

Thanks for your reply Bam.

Is there any (Maven?) way to superimpose the detection zone over the 12 second clip? I believe there I have a couple of examples where all green rectangles are outside the detection zone throughout the clip.

I will be very happy when Person (& Vehicle) detection has been “perfected” so that I will be able to expand my detection zones to include the shrubs blowing in the wind that trip the Motion Detection. When I reduce detection sensitivity to reduce the shrub movement detections I lose the Person Detections.

1 Like

I need to correct something here, I was messaged by another forum Maven and my description of what was happening is incorrect. so I tested my camera that I have seen most of my motion on sure that I was right in my description and indeed he was correct and apparently I have a camera that is not working correctly, especially because many of the other cameras behave differently then that one.

so the green motion detection boxes should only be working WITHIN the detection zone you set. so if you have a small detection zone in one area of your frame of view, that is the only area that should detect motion and trigger recording BUT, once it is recording, that entire frame of the video is used for detection of a person. so once triggered even if the person wasn’t in the detection zone, if they were in frame during recording, it should alert for person detection.

and thank you @Newshound for your catch there.

3 Likes

Interesting – I was just reading your FIRST description of how the motion detection worked, and I actually like that better. That’s not how mine seems to work, but that would be more useful, I think.

I’m curious how it works if a person is half-in and half-out of the detection zone, too. Obviously that would be a 3D effect and likely means that the person is very close to the camera. But I’m wondering if the person detection only pays attention to the cropped portion of the video or whatever. Haven’t really tested any of that, but it’s an interesting question.

1 Like

In my experience, and how I believe it works barring any malfunction, is if you set a detection zone for half the frame, then you will only be highlighted with a green box in that half of the frame.

Yeah, that’s how the highlighting works for me, sure. I was just reading @Bam’s FIRST (Inaccurate, I guess) description, and I kinda like it better. Haha. And like I said, it also got me curious about whether the person detection is only scanning the “watched” portion of the frame, when a person is half-in and half-out of the frame.

My understanding is the frame triggers motion detection, after which person detection looks at the entire frame.

1 Like

Is there a way to ask Wyze for a definitive answer? I would be curious.

Sorry, I shouldn’t have worded it so softly. That’s the way it works. I think that info either came out in the introductory video or with my conversations with employees, so I don’t have anything real solid to point at.

There is one sentence in the “Person Detection with Edge AI” document that at least implies it, though: “Wyze’s Edge AI is powered by Xnor.ai, and utilizes a machine-learning algorithm to determine whether frames within a motion event video clip contain a person.” So it refers to frames, not detection zone.

However, it is easy to test:

Set a Motion Detection Zone that is small or low. Using a tablet gives you greater control to set a small zone. Then walk thru it, triggering motion detection. You pick the portion of a body that enters the frame. Doubtful the AI could identify a person from a portion of a body, so that should indicate it is looking at the whole frame.

Obviously, the standard person detection rules would still apply – camera not mounted low, person from 2-20 feet, etc.

Thanks. I was mostly curious.

I have two cameras in the front of the house facing the sidewalk. I have the detection zones only set on my property. I haven’t detected people who stay on the sidewalk or walk down the street. If they walk into my yard or driveway, I am alerted 98% of the time.