V3 pro desperately needs a firmware update. Have not had one since they came out with it. The built in Ai detection makes way too many mistakes. A car is not a person! Please update already!!!
Isn’t the AI done on the server and not on the camera?
On the Pro, there is some AI in the camera.
There is currently a Beta in testing for the V3Pro in its second test round.
You are welcome to join the Beta to get that update and test it.
Most of the AI tags that are perceived as wrong are usually because a motion event creates an upload and the AI then identifies a different object, primarily a stationary one, as the tagged object. The car may be the motion that I initiated the upload, but something else in the image is being tagged as a person. The moving object is not always the one that gets tagged.
Since Event videos have no way of identifying the tagged AI object, there is really no way to identify with any certainty what the tagged object was. It comes down to a trial and error exercise with the DZ.
The Person Detection Edge AI on the V3Pro is very limited compared to that which is on the Server. They are designed to work together.
So what is the purpose of having on board ai if it’s inferior to the server ai?
I assume it’s because on-board AI is a lot faster, since it doesn’t have to push the data to the server to have AI applied and returned.
That said, it sounds like we are taking GREAT liberties with the term “AI” by which I assume is meant Artificial Intelligence (if it is being used to mean something else, my apologies and disregard). Basic (and largely inaccurate) pattern recognition is not Artificial Intelligence by any definition I’ve ever heard.
But it sounds good.
Correct. The implementation of the on-board Person EdgeAI in the V3Pro was intended to both decrease notification latency by only using the server based WyzeAI to confirm the cam’s positive tag and also to improve the accuracy of Person Detection tags by having a double check AI mechanism. Unfortunately, while the chip in the V3Pro is capable of running the EdgeAI, it is not powerful enough for the full package Wyze AI.
However they are applying the AI term, the software does have an object learning function, much like Facial Recognition AI, which Wyze is also using. Wyze has also been experimenting with user directed object learning AI.