It is NOT possible to set recording only on an AI event, using Cam Outdoor in battery mode this drains the battery. It is critical, for ALL cameras, to be able to eliminate constant recording when you only want event based recordings pre/post buffers. Say 15/30 seconds before and up to 1 minute post are usually useful.
Of course if installed SD Card then continuous recording would override.
One major difficulty with this request is that there is no local analysis, so the only way for it to at least initially know whether something is an AI event is to record the event first, then upload it to the server, then analyze it, then it knows yes or no. But it requires recording non-AI events in order to know whether they are an AI event or not, so since ALL events must be recorded in order to know if they are an AI event, then the only other option is to program it to DELETE non-AI events after they’ve already recorded, but this will only use up more battery.
Okay, but once it sends the first part of an event, it should know if there is a person in there or not, right? That’s true…but as long as there is ANY motion or any event at all, it needs to upload those new seconds too, because what happens if the initial motion was caused by a non-human, then 30 seconds later a human walks into the recording. Therefore the entire event must be recorded and uploaded to the AI in order to check whether an AI event is located anywhere during any of that event. 100% of that event must be recorded and analyzed by the AI in order for the AI to know whether it is an AI event at all.
That’s part of the struggle here. The AI processing is all done in the cloud after the video is recorded. What you are asking for would work a little better if Wyze ever gets around to adding local Edge AI onto some of the cameras. They mentioned they are looking into this, though if they do it, it will likely be limited to the V3, and maybe some future cams. Who knows though, maybe the WCO will get it eventually too. I just don’t see who they would make this work very well when they have to record 100% of all events in order to analyze whether they are AI in the first place. Do you have any ideas in mind for how to do this given their current constraints?
Well then dont delete the data but dont show it in the events. Also allow to set up recording option to NOT record all motion but any selected type of motion. Making a distinction between AI events and “other” events only confuses things. It is “events” regardless of the underlying logic.
All of these settings should existing in both recording and notifications. Because you might want to mix and match this would be cleaner more logical and function better.
Moving “plants” is a coming issue, this somehow needs to be excluded in a way that works… otherwise compromises the efficacy of the devices.
Yes this is why AI is being built into chips but the battery units even more, maybe include in those a onboard logic engine/chip.
So two issues, backed end technical requirements, which at this time may not be able to mitigate too much, and then user experience/app design. At least on the app design side that can be improved dramatically. Ultimately False positives have to stop.
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.