An update regarding Person Detection

That make no sense from a development perspective. If the only motion that’s of interest is ‘people’ why incur the overhead of managing (transmitting, storing, etc.) non-people motion only to suppress notifications.

If it is a notification filter as you suggest it should be clearly documented as such.

Only echoing what I believe I came across in another thread. Will paste the link if I rediscover it.

Are you implying that for there to be “motion”, there should be a “person”? That’s what this statement seems to be saying.

Yes, if person detection is turned on it refines the definition of motion as it must be identified as a ‘person’. If person detection is off than the ‘original’ definition of motion applies.

I can confirm that the person detection was run on already existing videos and it only enabled tags for people being detected and limited notifications. It didn’t reduce the number of videos but we’re interested in that in the future (though can’t make promises that this will happen).

2 Likes

You’re redefining what the options mean. And that’s not the way it happens in practice. Cite a Wyze statement, publication or document that says that.

Good to know. How I thought it works is incorrect. It would be more efficient from a data utilization perspective if it were a refinement of the what constitutes motion rather that being a notification filter.

It would! But it’s also harder and the risk of false positives or negatives carries greater weight. It is more tricky to do that with the hardware limitations of the device.

1 Like

It’s a moot point now that @UserCustomerGwen has confirmed that ‘person detection’ is nothing more than a notification filter versus a refinement of what constitutes motion.

I guess that begs the question of why it was decided to put the function in the firmware versus a server side function. I realize that server load has to be considered but if it can be successfully implemented in the F/W of a limited capability device the increased server load couldn’t be all that great.

I really don’t expect an answer but it would be an interesting discussion in which to participate.

If I recall correctly, it was faster to run in some ways (not all), security-conscious people appreciated the algorithm not happening on a cloud, and it didn’t require the cloud resources. I also know that we were hoping to be able to work it into a limiter someday. There are probably many other portions of this conversation but I wasn’t involved with them. These are just the things I heard about directly.

1 Like

Gwendolyn,

Question. Given that the new online subscription transfers videos from the Wyzecams to the cloud after it has been triggered, my question is, are there guidelines to the number of Wyze devices that can be on the a single residential internet service for a certain upload speeds, given an assumed coincidence transfer load factor for the Wyzecams?

For example my service is 20 Mbps upload rate. Live streams of Wyzecam feeds using either an iOS application or a PC Android emulator, is three (3), anything more and my feeds becomes unstable and all the cameras disconnect, one by one, until there is just one streaming camera feed. I currently have 11 Wyzecams with more on the way and I’m concerned with the growing overhead on the upload stream.

Reflecting on the old Wyze Maxdrive/NAS development - there were far wider reaching and stronger benefits for the user’s holistic home experience with that solution as customers continue to grow its use of IOT products and other family needs without spending more dollars for faster internet upload service speeds.

2 Likes

What does that mean exactly? It sounds intriguing. I know it’s moot for now since the feature in its current form is going away, but I’m still curious what you’re talking about.

We were hoping to develop person detection to eventually be able to reduce the number of Event videos in a way discussed previously in this thread. But that was a longer term project idea. :slight_smile:

1 Like

I dig long-term projects :sunglasses:

1 Like

Cool. Hopefully that will still come eventually. :slight_smile:

Any chance that the Wyze person dectect will be better. Maybe add pet detection?

I’m with you @masterep, exactly what I’ve been pointing out too. I’m not against their online/cloud services their offerings, but I hope they develop their products in a way that its still usable/functional offline on local Wifi/LAN. Internet had become widely available and getting cheaper, but as well as devices that consumes/using it, so on a regular household its easily getting congested. Your product price range targets common, middle, value and budget wise people, but its becoming a “catch” that it requires premium internet subscription, premium wifi router and maybe soon including cloud service will be required to for it to function properly. Same as @masterep; I think its better to have a hub / NAS / NVR for all the Wyze cams and smart things.

2 Likes

And also a newer device. My Samsung J3 is now showing significant video lag again. Made it difficult to position a V2 while hanging out a second story window and having to wait 20 seconds every time I moved the camera to a different FOV.
A $200 device upgrade solved the problem, but that is the equivalent of 10 V2’s.
It’s like buying an inexpensive item that requires an expensive infrastructure to be effective.

2 Likes

I second this.
For some camera I do want to be notified for dog, for another not. The current implementation set only to detect a person, always triggers when my dog passes by.