The Pixel algorithm is not meant for outdoors

Cameras
1

I have 5 v3’s that cover the outside of my property .

The pixel algorithm is not totally meant for outdoor use . As we all know , the cameras run off the algorithm to trigger an event. My cameras are constantly being triggered by light changes , shadows , tress waving , bugs , birds , wind , cars headlights , lights reflecting off of objects , rain droplets , wind , etc

I have a v3 that’s mounted to my window overlooking my front porch and front yard . It also has my street in frame so it’s constantly being triggered by the cars headlights at night .

I have tried blocking the roads but I would have to block out the whole frame and whenever I walk in frame it is not able to determine if I’m a person or motion so the AI becomes useless .

2

What version of the Firmware are you at on your v3? Also what is the Version of the App you are running?

I have found after an update, simply turning off all toggles in Notification and Event Recording and then removing Cam Plus, and then logout of the app.

Then Log Back in, Assign Cam Plus, go to Event Recording and turn it off as it will be activated when you add Cam plus. This sometimes causes an issue when you want to assign notifications. Go back to the Live Stream and then simply go back to Event Recording and turn on all Event Recording options you would like especially in AI, then go to Notification and turn on all Notifications you want.

I don’t have an issue like this, but I only alert on Person, Pet, and Package. My Driveway Cam faces the streat and cars pass all the time, so I don’t alert on Vehicle.

Something to try and see if it helps

3

Hi Spam

V3 FW : 4.36.9.139

App version: 2.33.0 (17)

I have turned off vehicle , bc whenever a light change or other motion events were in frame bc of my parked car it would notify me every time of a vehicle event.

The issue I’m having is mainly with the detection zone, the placement of the camera has my street in frame and it motion tags and gets triggered by cars and their headlights all the time.

I don’t want this to clog up my feed so Ive blocked out practically the whole frame and whenever a person walks in frame since the whole frame is blocked out the AI has trouble labeling it a person events .

Which is why I’m so against the rigid boxes because they cause more harm than good , pull and drag will be more effective and accurate !

4

The detection zone used to be pull and drag, nobody liked that either :grin:

5

I understand what you mean with the detection zone. I have used other systems which allowed free flow of detection areas and not blocked. Worked nicely.

Thought you were indicating that notifications have stopped. Question, since you have the camera in the window, do you think the glare on the window could be adding to the issues? Have you tested with the camera outside to see?

6

But it was box driven, correct?

I have used some with 8+ points and you drag them where you want them. You could create a star of sorts, octagon, or whatever you wanted to denote the detection zone.

7

Yes it was just one big box, you couldn’t make separate little shapes or boxes.

1 Like
8

Yea, that was not the best, but I could build what I wanted. I think you could have up to 3, if I recall correctly. Been some time now. :slight_smile:

9

I think it might be a misunderstanding on your part Spam . I have a v3 that covers my front yard and my street is in the frame so I’m order for me to reduce the camera motion tagging and getting triggered by every single car and headlight . I have to block out the road and i end up having to block the whole frame which reduces the AI to work properly and tag events . So when I block the whole frame , when a person walks in frame , it’ll just label it as motion bc of the pixel blockage and not having enough pixels to correctly determine and label the event

1 Like
10

Yikes , who didn’t like the pull and drag . They must be maniacs , it’s so much more accurate

11

Ring has pull and drag detection zones with 3 zones you can add on one screen. With the ability to make shapes and sizes to your liking

This is exactly what we need on wyze cameras

12

Got it.

13

If I remember correctly it was just ONE big box that you could pull in from the sides and from top and bottom. I don’t think it allowed you to make different little boxes, I don’t think it did :thinking:

1 Like
14

Right , I saw that appear one time in my detection zone when I opened it up . Instead of the boxes it was a pull and drag . It just showed one whole box

Ring has the option to add 3 zones that you can pull and drag to make them the shape and size you want . This should definitely be implemented to the wyze cameras . The rigid boxes suck

15

100% agree. I get “Person” detections when it is just rain in the 12 second clip. I do submit them so hopefully they are working on improvements.

1 Like
16

I was having the same issue, but I also reduced the sensitivity of the detection to about 40 and it still sees people on the street. I turned off vehicle detection and I no longer have vehicle notifications. I do still get person detection with Alexa.

1 Like
17

I did not think that was the way it worked. I thought it looked at the whole frame for determining humans…

18

The ai used to scan the whole frame, but this was fixed recently so now if a person is at least a tiny bit inside the detection zone it will identify it even if the rest is outside.

1 Like
19

Yes exactly, but that is NOT what @Rulwiz said - he or she is saying / thinking that a person is not identified if only an arm or leg makes it into the detection zone.

20

If an arm is inside the zone but the rest is not, it will use the whole human to identify as far as I know.