As the current mechanism, it usually takes 8 minutes to update the status from on to offline, and 5 minutes from offline to on. Hope can explain what you ran into. Thank you.
I’ve the Android app 2.3.80 and FW 4.9.4.37 and time lapse is broke. I’m about to flash back to FW 4.9.4.28. Did the FW update do anything to fix the time lapse problem ?
@WyzeRoy in the beta version of the vehicle/person detection (as opposed to alpha FW), did you get rid of the lower left corner showing feedback of what the AI thinks it is seeing? im getting generic motion triggers with no clarification in the notification and it is not showing the readout in the lower left
I have noticed in the “Events” tab/button that it can sometimes correctly identify the video as “Person detected”, “Vehicle detected”, or “Motion” without any tagging. If I happen to walk by the camera very quickly it will tag the video as “Motion” instead of “Person detected”. Of course I will then correctly tag the video but there seems to be something odd about how it detects people movement versus motion. Apparently all you have to do is run by the camera and it will tag the video as “Motion” instead of “Person detected”.
It’s pretty apparent that it’s doing 2 things at this point - detection motion through pixel movement and radiated heat via ir. I suspect the ‘Recognition’ capability will require a lot of work but the I.d. buttons will give them feedback to eventually start training an algorithm. If you’re not giving off heat for long enough as you walk past for it to recognize the thermal change it will identify as motion from the pixel movement/changes.
Thanks #Solardave that can explain how it works a bit. I am concerned for how it will be able to use IR radiated heat detection in a Texas climate with the sun heating up the concrete underneath the camera. Lots of radiated heat coming from hot surfaces in the summer all the time (especially vehicles recently turned off sitting in the driveway) and if there are swaying tree shadows (for the pixel changes) within the cameras view at the same time, it will likely be detecting people/vehicles all day long. I hope it is using another layer of checking for another mysterious condition to help with a situation like that. Perhaps this is where XNOR’s AI solution helps for the recognition part. Thanks for the insight.
I am so happy that this feature is not just for newer camera hardware, I was not expecting that from Wyze but kudos to them for making it so. I thought they were going to do this part themselves but it looks like they may have quickly figured out how complicated it is and decided to partner with someone who already had the solution and could scale it for them efficiently. Win win for everyone but I hope it was not too expensive, I like Wyze’s prices!