All the more reason to switch back to a local Edge system, even for investors’ profits’ sake.
I could be off by several orders of magnitude but I would not be shocked if the AWS bill was in that range. “The cloud” isn’t cheap at scale, Medium sized footprints easily hit $6M/year for a thousand or so compute nodes - it’s cost effective vs leasing a quarter of a data center, but it’s not free.
This is an old post from back in May, but I thought it shed light on the whole edge vs cloud AI discussion
So… They imply they could do edge again now… Maybe only for those that don’t do CMC or have a sense bridge using up the processor.
Hopefully that’s what the V3 is all about…2K with improved processor and all local as a possibility.
I really hope so!!
Person detection is not an easy algorithm for that little $20 camera to handle, especially in real time, then the camera cpu has to do recording, streaming, notification, among other things. There is not an Intel I7 cpu in there like your computer. I believe people are starting to expect too much from a $20 camera or Wyze is overly optimistic/promised on how many features it can build into it.
Personally, I put in a sd card, tell it to record all motion and then call it the day. I paid about $10 each for 64gb u3 microsd card. I just wish there is an easier way to extract the recordings from the sd card to review on a computer. Currently, I do a fast review of the recording from the app. If anything needs to be further looked at, I pull the sd card out and put into my computer. Yeah, since I don’t use it as a real security camera, they are all put in places that are easily accessible.
Does the bridge actually use the processor or just the Wifi and get power from USB?
It’s done it in the past.
The bridge uses very little of the main Cam’s cpu.
We know this, because the bridge can be used on a non-Cam (say PC, RaspPi, etc), and it was discovered there was a extremely simple serial protocol being sent back and forth.
We also know this, because the protocol has been reverse engineered.
Like you suggest, the bridge uses the Cam to get power to the zigbee (like) radio in the bridge, as well as a way to leverage the Wifi in the cam as a way to get the sensor data to the Cloud.
It was an very smart way to do things. What better way to lower costs, than to use the Cam to hijack the resources you need?
If anything, I am extremely disappointed that they didn’t do more with the Bridge and Sensors.
Back when they first came out, they sure had us believing that many other sensors were following. Water sensors, Temp sensors, Light sensors…
The sky was the limit!
I just don’t know happened.
Did they determine the viability of the sensors just wasn’t worth it? That a Scale and Band had a better profit margin, with higher growth potential?
Did the person that ran the Bridge/Sensor division piss off the wrong people at Wyze, and the project got bumped down to the level of Milton from OfficeSpace?
Would love to have been a fly on the wall in the offices to have heard what their rational was for minimizing the Sensor group.
Yes it was but still not to many people’s expectation. The AI company sold it off to Apple and I don’t believe Apple plans on using it on a $20 device. Besides, trying to reverse engineer the feature and at the same time avoiding patent lawsuit costs a lot of effort and money.
Good info - thanks!
Never underestimate the power of a red stapler!
I don’t believe Apple believes in $20 devices.
I’d love to add more types of sensors to my house!
Crazy how it makes pan cams spaz out then. I tried it on mine and it basically overheated and acted all weird and lots of malfunctions. Then I read online of others with similar problems and that you should never connect it to a pan cam because somehow it overloads it and causes lots of problems. I tried 2. So I took the advice and switch to V2’s, and it worked much better!
You might say something about how it’s just draining power, not the processor, but I Daisy-chained a V2 to the pan cam (connected a V2 into the pan cam’s USB slot to fully power another cam) and it works perfectly fine. If it was just a power issue, then the V2 should be stealing more power than a bridge and cause equal or worse issues, but the bridge overheats it and the V2 doesn’t. So it’s definitely not just a power issue…
Could be a firmware conflict with the pan cam, though I’d think it would be using the same subroutine as a V2 for that part of it. All I know is bridges are a bad idea with the pan cam and better with V2’s, and it sure sounded and felt like it was overwhelming the processor, but I suppose that is hard to actually prove.
Well, considering the primary business plan is to just rebrand other developed products with a few modifications, maybe it was hard to find all those kinds of different sensors that would all play nice together in the same way without tons of new coding or something?
We’re already feeling the ramifications of this situation, and we’re NOT liking it,
First - ALL of our person detection simply just stopped working altogether about 1 & ½ weeks ago. What’s with this???
We’ve been using the Ai… person detection… whatever since it was introduced.
(1) Are you calling the “beta” version the one Wyze had started developing?
(2) I don’t know if we started it up on every one of our cameras, but was our intention. I don’t know if that means we have it on all our cameras, the ones associated with our account (email address) or what?
(3) And what I really want to know is… what happened to our person detection… NOW. All in the last week or two - it’s no longer working on any cameras, especially and specifically on ones we KNOW we had it set to use. (door entrances) I thought it wasn’t supposed to end until September?
This isn’t looking well at all for at least our relationship with Wyze. First having to pay for extra time… with complete motion. And now person detection potentially be fee based after your promise?? … which then I can see - you may attach a further fee of adding complete motion to person detection? Is that the plan??
Additionally - this all doesn’t even mention anything about how all of this will affect any future Wyze products!! Including just a few of the new recent products you have some-thirty products in the works.
What fees will we be looking at?
are all these fees things pertaining to if you use cloud storage only? Or are they talking about paying a fee to use the Person Detecting feature, regardless of cloud or sd card recording?
Person Detection only works on cloud recordings. It does not work on SD card recording.
So, Apple purchased Xnor.ai for $200 million.
If it was as easy as you described Apple could have done it themselves but instead paid $200 million.
that really sucks. i will stick to my original way to record all motion on sd card.
Except that according to the spec sheet linked earlier, human detection already was a feature of the SoC that the camera is built around.
Also, apparently that firmware rewrite was AFTER Xnor had already been running fine on the cameras. They made more room but didn’t use it for local PD.
(“Edge” is unnecessary. The terms “local” and “client” already have perfectly fine acceptance for the same purpose.)