What is the logic Wyze used when making the AI label vehicles that are parked, instead of moving? Why would anyone want to know this?
2 Likes
The majority of video AI (not just Wyze specifically) has traditionally been trained to analyze the frame/picture for objects in each individual frame of a video (which we know is just a bunch of single pictures changed rapidly). So this is actually the normal way Video AI detection has usually worked in most cases. So it’s less about logic than the way video analysis is typically done.
In order to ignore parked vehicles and only tag moving vehicles, the programming has to be entirely rewritten with a different algorithm that can take multiple frames into account, and compare differences between them. So it still has to analyze individual frames as normal, but then it needs to also compare previous frames to subsequent frames and determine an algorithm for what counts as enough movement between the 2 (or more) stationary frames to count as sufficient movement, and then pass that on. This effort uses a lot more server power and costs more to do, so it will also increase costs in multiple ways.
That’s not to say it’s not worth doing. Several large companies have also started to research or implement this moving object detection feature because it adds a lot of value. Wyze has also been working on this for a while and recently stated that they now have a functional version of moving object detection (like only moving vehicles, not parked ones) set up and is now being tested.
They have also updated the wishlist for “Objects in motion (not stationary) notifications only” into a status of “In Progress” and “Researching”.
So they are working on it, but it is a lot more complicated to implement from scratch than people think. I really look forward to it though. I have personally turned off Vehicle notifications on the majority of my cameras for now because I also don’t want parked vehicle notifications nonstop. So I am very anxious for that update and more than willing to help test it when/if it graduates from Internal testing to beta testing.
7 Likes