they do differently than what u might think, i guess.
instead of tacking the problem directly, they will implement indirectly by detecting (1) person; (2) vehicle; (3 -maybe later) pet. u can choose when u want to be notified - person/vehicle while ignoring others.
there is beta test on detecting person, vehicle, person+vehicle, the detection of person is very accurate whereas vehicle less so. once thy sort out this, it is easy to implement on app.
We have multiple Wyze Cams. Both indoor and out. All work amazing with the exception of windy days and they chase anything that moves. Amazing products for the money.
We have everything Wyze makes but light bulbs and we are on the waiting list for those. Canāt wait for them to go public. We already have cash set aside for stocks.
I just noticed that this topic says ālaunched.ā When did that happen exactly, and what was launched? Just curious for details. Should we notice fewer false positives since a certain point?
It would be great if this can be ongoing, similar to the āperson detectionā feature. Teaching the camera how to ignore false positives, basically.
Oh, I see. I guess those seem pretty separate to me. āPerson detectionā isnāt the same thing as āalgorithm to ignore blowing leaves/shadows.ā That wasnāt totally straightforward to me. Haha.
I wish the algorithm was able to recognize bugs flyingā¦ we got so many events triggered by moths and other flying bugs passing in front of the camera and reflecting the IR light which then triggers the event recording and the notification of false positives
The algorithm doesnāt have to be fancy at all. Just donāt notify when the movement is for less than two seconds (as mentioned several times in this thread). If someone wants to call this āartificial intelligenceā then fine, call it that.
Cool. As an incentive for you guys to implement this, it should cut down on your cloud storage and bandwidth costs, since a lot of false alerts wonāt need to get uploaded in the first place. (Ignoring very short motion events is also tied to this.)
Maybe that savings would allow you guys to implement a separate cooldown period for person detection. Thatās really the main feature Iām waiting for ā reliable person detection notifications that are independent on whether the camera got triggered by something else in the moments before a person appeared.
Person detection is run on existing videos so thatās pretty difficult to run. Itās more of a tag on a motion detection video than a separate video trigger.
I was told somewhere that person detection happens on the camera, not in the cloud. Is that not accurate? If itās happening on the camera, I donāt completely understand what the difficulty is.
I didnāt understand this sentence at all. Haha. Sorry, can you rephrase it? Itās not computing in my head for some reason.
Also, I realize weāre steering this thread a little off-topic. Sorry. We could continue in a private message if that makes more sense.
It is happening on the camera. But the 5 minute cooldown happens on the camera as well. The video is recorded and in the cameraās memory when the algorithm is run on it. So the cooldown has already started when the video is being analyzed.
Motion Detection Video > Person Detection > Stays labeled as a motion video or becomes labeled as a person detection video > Uploaded to the cloud > Notification sent to user
If you want to continue from here, we should work in a message.
I am actually also interested in finding out more about it since I have had 2 consecutive motion events one sound another person on the same camera that were just 2 minutes apartā¦ so 5 is a sort of a āmaxā and perhaps cool down could be anywhere in between? Thanks @UserCustomerGwen
When you look at the 5-minute cool down that does not apply to sound events, just motion initiated events from the camera.
So theoretically you could have it set to trigger for sound, motion from cam, and motion from a motion sensor, the only 5-minute gap would be between motion events detected by the camera motion sensor (this includes person since it is a motion event from the camera)
Makes senseā¦ so one could completely forgo the motion detection from the camera and use the one provided by the motion sensor insteadā¦ would that allow person detection? (Probably not since that needs to analyze recorded video)
I guess the motion sensor is a motion sensor with IR? And only usable to trigger video recording upon detecting a motion, but thereād be no ārecognitionā of Persons or areas of motions in the video like it happens on the camera.
I wish I could buy single camera receivers or that they were an optional add on available when ordering the camera itself (perhaps the micro SD too should be an additional option that customers could order when purchasing a camera) I should perhaps add this to the #wishlist