I am slightly confused regarding motion types with people/vehicle detection. There are scenarios such as people in motion with stationary vehicle, vehicle in motion with stationary people or both. What about pets. They are smaller than people, I have cameras set up to monitor my pets. Crawling infants are people and similar in size to your average cat or dog. I would not suspect detecting crawling infants as a detection scenario . My point is, there should be an option in the beta to select not only person or vehicle but to add pet as an option.
I have another camera that tags for all three so that you can set notifications based off what comes into the area. This is all still new to the them as they are getting the algorithm done for the main bit. Persons and cars. They will probably add animals later. Give it time. So far its working wonders.
I agree, I have had fewer bug detections than in the past.
I posted this in a different thread here so I figured I’d copy/paste it here as well because it apples here as well… couldn’t figure out how to copy link to thread… have to find a young person to help!
With p&v tagging when I send feedback… I have a vehicle parked in my driveway and a person (me) walks up the driveway how should that be categorized? Person and vehicle because there is a vehicle in the picture… also, at night I live on a corner and headlights set it off even if the vehicle is not in frame… how should that be categorized?
Wow, wish I was a Beta tester for this. I am going to try and get on board.
I have the same question.
Should we tag videos by what triggered the motion or by what objects are present in the frame? I have many videos that were marked “vehicle detected” where it was a bird or insect that actually triggered the motion alert but a stationary vehicle is in the frame.
I hope the training goes well. I hope it leads to not only recognition, but reverse recognition. Because it would be nice to get a notification when the motion is a person, but not when the motion is a vehicle. I live on a busy street.
Wyze reps say tag whatever is in the frame, even if it’s stationary.
Will be interesting to see how sophisticated Wyze can get with this. Distinguishing People/Vehicles against birds and insects might be getting at their limit and needing a PIR motion sensor working in conjunction with the camera BUT I hope they can create some magic to make it work better for situations like this regardless.
#prjct92eh2, I am confused. So, if my parked car has been within the camera view for several hours and I walk up to my parked car then we are supposed to select “Person & Vehicle” not just “Person” even though a person caused the recording event? If that is the case my driveway camera will almost always include both a person and a vehicle. Most people are more likely to think to tag what actually causes the recorded event not what is in the video…just need some clarification, I wish a moderator would step in and clarify this.
That is correct. Wyze staff @WyzeLi clarified that to us during alpha testing.
This testing is to improve their algorithm but many of my triggers are birds on the ground, flying through detection zone and dogs. Maybe add those as choices in the test scans.
I asked similar questions in Alpha and the Wyze team said you tag whatever is in the frame, even if it didn’t trigger the motion and isn’t moving. I personally find that less useful. I too have a scenario where two of my cameras look out over the driveway and therefore always have my car in view, but it’s rarely moving. I would prefer to be alerted to what moved or triggered the alert, but this is work in progress and all good stuff.
As for the “pet” questions,… as long as you don’t have a Pet Human, then you shouldn’t tag a pet as a human or a vehicle. It’s just general motion (so pick none). I can see why you would want animal detection, although I presume that is a lot more complex for the algorithm to identify. A feature request for another day I think.
its pretty simple, its AI, so its constantly learning by example, by tagging everything in the picture you’re giving the AI more information and variations on those items to compare to, the more it has to compare, the better it will get, and the less errors you’ll have over time. that is why WYZE is telling you to tag everything, its not about being useful for you, its useful for the AI.