Wyze Anything Recognition - AWESOME AI updates coming for Cam Plus

Services & Integrations
1

Wyze announced they are working toward launching a program called:

“WYZE ANYTHING RECOGNITION”

In short, each individual user will be able to train the AI to recognize ANYTHING we want it to. In the Reddit AMA they also clarified that they have plans to “bring the new model everywhere: Wyze services, devices, etc… in a backend agnostic way and each person will have their own set of models that can be shared. We start small, storing the model in the server first.”

For examples of what you’ll be able to train the AI to do for yourself, watch the video from about 20:23 through 23:40

THAT’S AWESOME!!! It’s virtually limitless. We can [hopefully in the near future] train the AI to recognize any object or nearly any situation, and have those custom detections and alerts specific to our own personal needs!

And we can share that training with others to be able to use too. I am going to train the AI to recognize the difference between when a person is using my swimming pool/in the pool vs not in the pool, and only send me an alert if someone is in the pool so I can make sure they’re okay and there is someone responsible and old enough supervising.

More stuff mentioned about this in the Reddit AMA:

From WyzeShawn
“…it is possible to learn a good customized model for each user & camera with much fewer data due to the recent process of few-shot learning and AutoML.TL;DR version, this time, we want to bring the cutting edge AI technologies more approachable to Wyze customers, and we want you to be able to train your Wyze Cam to recognize things. Stay tuned with us later this year! Something big is coming. :wink:”

I don’t know about you, but with this new revelation about how each individual will be able to create custom detection models for ANYTHING WE WANT, I am super excited and DEFINITELY more than willing to fork over $99/yr for Cam Plus Unlimited (99 licenses) since I can train my cams to recognize and alert to nearly ANYTHING I want for each individual cam.

I was previously thinking about switching some cams to RTSP and running them through Home Assistant plug-ins to be able to do this exact thing (create my own detection training), but if Wyze is going to allow me to do that all with their AI with unlimited cams for just $99/yr ($8.25/mo avg), I am totally sold with sticking with Wyze AI! That’s all incredible news and will be SLAUGHTERING what the competition offers IMO. I am hoping it works out like they’re hoping and planning and I am really excited after getting some of these clarifications of their vision of where they are trying to take all of this, aren’t you?

What kinds of special things would you train the Wyze Anything Recognition AI to recognize or alert you to [if it is capable]?

12 Likes
2

I’d train mine to watch for Christie Brinkley walking up my sidewalk. Hello !!

3 Likes
3

I would train mine to recognize the car that’s been parked for months in my driveway doesn’t have to be recognized as a vehicle every 15 minutes.

7 Likes
4

Since you brought it up here, I actually asked them in the AMA if they would work on detecting a difference between a moving vehicle vs a stationary vehicles so we could have it ignore stationary vehicles just like you’re saying you want to do. WyzeShawn replied:

Thanks for sharing this feedback! Yes, this one has been in our backlog from day 1 of when we launched the vehicle detection, actually. We totally understand how this is useful, and the team is looking into a solution in a broader scenario right now. We’ll keep you all posted!

So Wyze is looking to solve this issue for everyone. That will be nice!

6 Likes
5

I would like to add myself to the list of people who would sign up for $99 a year Cam+ covering 99 licenses

1 Like
6

I was just thinking, when this comes out I’m going to have a camera dedicated to recognizing the difference between when my stoves and ovens (I have one on each floor of the house) are on and off. I will train the cam to tell when the dials are turned on or off, and maybe warn me when it has been on for a certain amount of time, or if the stove was left on, without a pot being on there, etc. Or maybe it can see a pot that is empty on the stove while the stove is turned on, or any number of other risky situations similar to those.

This came to mind because I started to cook a pot of ramen, then got distracted waiting for it to boil, and forgot about it until all the water was evaporated! Luckily I remembered before it got too bad. There have been times when people people have forgotten to turn off the stove or oven though. A neighbor accidentally burned down their house this way.

For me, it would definitely be worth the cost of dedicating a camera to watching the stove/oven and alerting me to certain situations related to that. I acknowledge it’s still my responsibility not to be reliant on such alerts, but they can be something that would be helpful and put my mind at ease that there is potentially something that could help. Maybe when it is actually released we’ll have to have a whole sub topic on tips and tricks for implementing all these ideas that people might be interested in.

2 Likes
7
8

Yeah, I have also seen people use contact sensors to do this when they have nobs. The problem is that not everything uses nobs anymore, especially ovens. Lots of those are buttons. However, if we can train the camera to recognize what the display looks like when something off compared to on, then each individual can train the AI for their unique device without Wyze having to create a few hundred variations of things just to fit everyone’s different iterations. I love that I’ll be able to train it to look for exactly what I want for my needs.

1 Like
9

While you are waiting on AI to learn about heat on your stove and over, PLEASE, set a timer/reminder on your phone or smart home assistant to remind you that you left the stove on upstairs. We do NOT want you to burn your house down. We need you here answering questions, sir.

2 Likes
10

Thanks, that’s good advice. I have usually only used timers for things when they say something like “cook/boil for X minutes” but never thought about using them for something like when I’m just waiting for the water to boil, or cooking something until it looks done or whatever…but I should do this now! I do have Google and Alexa devices throughout my whole house, so I think I will start to do as you said, tell the assistant to set a timer or reminder for a certain interval when I use the stove/oven, then just tell the assistant to cancel the timer/reminder after I’m done with it (otherwise it will remind me to check). Makes a lot of sense and will be really easy. Thanks for the suggestion and concern.

I still look forward to the AI recognition option since I have a teenage daughter who is often…not very mindful. :slight_smile: So it will be good to have as an extra precaution.

1 Like
11

Good, I just wanted to do what I could to save your house burning down, like the someone you know. Thanks for your reply. You are one of the good/goodest guys. :slight_smile:

1 Like
12

I thought that was what the V2 was made for.
I’ve set one looking over stuff that took a while to bring up to a slow boil on low heat and then got the notification on my watch when it detected the motion.
Was fairly kludgy though and not really practical.
Not as good as watching Trojan Room coffee pot.

1 Like
14

I don’t see why not. You would just have to train it with a few examples of lights on vs off (in day and night time) so it can tell the difference. The more examples, the more accurate it will be.

2 Likes
16

They have not stated any ETA other than that they are working on it and it will come to other models whenever they can.

The problem with giving ETAs is that people get entitled and take them as binding promises and it ends up causing more problems, so a lot of companies, including Wyze, learn to stop giving ETAs because it often hurts them more than it ever helps. Much better for them to just say it’s on their roadmap and leave it at that.

2 Likes
17

Carver, as you know, some items have been in process for four years don’t you think that’s a bit of a stretch?

1 Like
18

Which things in particular? I know some things have been requested forever, though that does not necessarily mean that Wyze committed to them back when users initially asked for them.

For example, Dark Mode and Landscape Mode are not something that can be done quickly and simply. They will require a full rewrite of the app. And while users will say that Wyze has been working on these for X years, that is not exactly accurate. Though Wyze does intend to get to it, they have not been on the roadmap, nor given an ETA. Even the actual roadmap itself is not a guarantee of something. Both the Wishlist and the roadmap have several phases, one of the most basic being the “researching” phase, which means they are actively looking into options for implementation. It does not mean they’ve promised to do it, but the various teams are considering everything about it. What supplier can they partner with, how much will it cost, how likely is it to sell at the required price, how big is the demand, how many people would they need to hire for it, how much is the testing going to cost, and probably countless other things. A lot of stuff that reaches the “Researching Phase” probably doesn’t pan out. For example, in one AMA they said they were researching offering Smart Pet tags, and in a later one told us that things didn’t work out with the supplier as they needed and that project couldn’t move forward. Sometimes during internal testing something turns out not to be feasible for implementation and gets vetoed/canceled. There are also often unforeseen circumstances like the economy shutdown which ruined several of their projects. They had to pull teams off several projects they were working on and re-assign them to other critical things to keep up. I know that they were previously working on the 3rd version of the app before Covid, and had to pull people off of that for a while. They are now trying to hire more software/app developers again and they may eventually be able to spare people to start working on it again.

I do know that they are always actively working on things on the roadmap, and it’s easy to see several of the things they’ve implemented based on user requests by running a search on the #launched tag which shows about 152 user requested items implemented, not including all the Rules that have been implemented by request.

https://forums.wyze.com/search?q=%23launched

I guess, to me, having worked with software and product engineers in the past, the “Wyze working on something” is not just a black and white thing. There are several points along the way that can make a big difference in whether development is successful. Most companies learn really quickly not to tell anyone about anything until everything is basically 100% done and ready because otherwise everyone gets mad at them at how long it’s taking and when things change or cancel, etc. So companies become less and less transparent and more secretive. I’d hate for Wyze keep moving farther and farther that way to become like everyone else. I love when they share their roadmap even if things change for countless reasons. I’ll take that any day over total silence on everything 100% of the time until something is finally totally completed and ready.

Basically, I buy a product now for it’s current functionality at the moment I bought it. If I don’t like it how it is now, I return it. I avoid buying anything and keeping it in the hopes that it might one day get something added to it later and feel entitled and angry when/if that never comes. That just mucks up my life and quality of living. That’s not to say that I don’t hope for and push for updates and new things and support things, but if they never happen, it’s not going to really upset me too much because I bought and kept the thing for what it already was, not what it could or would eventually be. The non-entitled perspective is a lot happier way for me to live.

1 Like