I have 9 V3 cams. One of them detects a person on nearly every recorded event. All cams have a 64 GB SD card and all are set to record continuously. Only this one cam seems to mistake vehicles, insects, lighting changes, trees waving, etc. for persons. It also does properly detect actual persons, which is a good thing.
Any ideas? I have not had time to do much troubleshooting because the camera in question is at my dad’s house which is some distance away, but my gut tells me just resetting the camera or deleting it and reinstalling it may not do much. Definitely worth a try though. Any other thoughts? Appreciate the help.
I suspect that the cam isn’t mistaking those objects as a person. Rather, those objects are moving, therefore initiating a Motion Event Video Recording. The Video is being sent to the AI Bot on the server for AI interrogation and SOMETHING within the Active Detection Zone is being interpreted as a person.
The object moving in the video is NOT always the object tagged with the AI. It is only providing the motion necessary for activation of a video upload. The AI looks for ANYTHING in the frame and unrestricted by a Detection Zone when it looks to tag objects. Motion is NOT a requirement for tagging. All objects are considered, moving or stationary.
If there is a tree, mailbox, sign, shadow, bush, fountain, chair, etc., that might look like a person… That is what is being tagged.
I had one V3 doing this. Drove me nuts until I learned why. I started using the DZ to block out one single object at a time until suddenly it stopped cold. It was my mailbox that has a girthy bush in front of it. That was my phantom person.
This is very helpful. I’m going to try your idea of putting detections on blocking around certain objects within the field of view and see what happens. Thanks very much for this idea!
I blocked all the items you marked except for the tree trunk and 99% of the false positives for person detection have stopped. Great solution, thanks again!
