Smart Vision

Whatever happened to Smart Vision? I hadn’t had a use for it for a bit, and just realized it disappeared when I went to set one up now that I have a good use for it. Kinda bummed.

I think they just barely removed it within the last update or two. At least for all of us non-cam protect subscribers. I’m not confident if cam protect still has it or not.

I noticed it go missing as soon as they added the wise intelligence tab to the app.

1 Like

I hope they bring it back.

1 Like

What is Smart Vision?

1 Like

It is/was a feature where you could train wyze to alert you for a very specific thing (like a door being open, or a food bowl being empty). You would take a bunch of photos of the door being open, and a bunch where it’s closed, and then you could get specialized notifications for that specific thing.

I had originally used it to alert me to my elderly cats litter box use (2 of them were having some issues late in life), but I hadn’t had a use case for it since they passed in 2023).

1 Like

That seems like a cool feature, wonder why would they remove it :thinking:

2 Likes

No subscriptions…

So no revenue for Wyze, still they could’ve ported to Cam Plus/Unlimited.

1 Like

It was VERY clunky and didn’t make sense as to what you would do with it. Unfortunately; as it is in many parts of the app, the English was terrible too so it was pretty doomed to fail. Half-baked overall.

For example, this dumpster fire of a sentence (below) has been inside the “Feature Tips” of the Thermostat since it’s implementation. Half-baked and then they just seem to move on and not finish anything. cough Dark Mode

I thought smart vision was very user friendly, personally. The instructions for setup were clear and it worked well.

1 Like

I could’ve been over thinking it but I felt like they were confused. Probably me making it too hard.

cough, cough Landscape Mode :wink:

There were a lot of steps, and you had to click train once you had your sample photos.

It also only worked for basic things like a door open or closed, a cat present or not present. I tried to train it for a specific cat, to exclude the others, but that didn’t work, as well as training for birds and not Squirrels.

The idea was good for specific use cases, and you could change the verbiage of the notification you received.

The only major qualm I had with it was that you HAD to leave the photos on your phone that were used for the training. And they were stored in the camera roll, rather than a Wyze folder, so it was a major frustration for Android users. I had made the suggestion to change that to at least put them in a different folder… But it went nowhere.

Actually smart vision required a subscription. The open public pilot beta required Cam Plus. Then they restricted it to Cam Protect, But grandfathered in all the people who were on the pilot beta. I don’t know if it’s still available with cam protect or not, but I can confirm they took it away from the rest of us. But regardless, it definitely always required a subscription.

It was a pretty cool feature though. It just took a lot of patience to complete the necessary controls and detection Screenshot examples For everything that would count as a detection and then lots of examples of things that do not count as a detection. Then it stored those examples in the cloud under your account for the AI to check. Then if the AI saw frame that was more similar to a detection example, then a non-detection example it would tag and notify of whichever one you asked it to notify you about.

So, for example, if you wanted it to send you a notification anytime your pet’s food bowl was empty, you would take a bunch of pictures of what an empty bowl looks like, and then a bunch of pictures of the opposite. You would also want to take pictures with and without your pet present, take some in the daytime in good light and the night time and night vision and almost any other kind of difference that there could be. The more examples the better. So it took a little bit of work to train the AI, but it was a pretty cool feature idea. People could set up detections for almost anything they can think of in a still fine frame.

It definitely wasn’t an issue over subscription though, cuz it was always part of a subscription. It had to use the cloud AI in order to see if an event had the detection or non-detection similarity.

I think the main reason they got rid of it is that hardly anybody ever used it because they’re too lazy to train their own Detections or they would just set up one bad example photo and then wonder why there were false Detections.

I’m bummed. Did they ever fix where the photos were stored before they just took it away? The last time I used it the photos were stored on the device (and not in a separate folder but in the actual camera roll). Once trained if you deleted the photos from your phone it stopped working properly.

1 Like

I’m not sure. I rarely used it. It was too much work. :sweat_smile:

1 Like

You people have to stop taking my words out of my mouth. I was going to say that modern humans are lazy by nature and this seems like way to much work, at least for me :rofl:

1 Like

Nothing wrong with speaking the truth and that’s all you did. Beat me to the punch :wink:

1 Like

The thing I loved the most about this was actually getting hands on experience to how AI recognition training works and how complex it really is. It was a really cool learning experience with trial and error and figuring out what helps, what I forgot to take into account, what I could do better, etc. It gave me a great appreciation for the effort it takes to train the AI in different circumstances.

But yeah, definitely too much work to do it too often. Most people wouldn’t bother, and those who do often don’t do it right and then get mad their detection isn’t perfect or working magically the way they envisioned with just 2 pictures.

1 Like

I completely agree!

1 Like