Detection option based on detecting differences from what is normally in the frame (AI Learning)

Would it be possible to add an AI detection option that would learn what is “normally” in the camera’s view and only tag movement when it sees something out of this norm? Could cut down on notifications for trees, waving flags, blowing grass, vehicle tags for personal cars parked in the driveway etc. I’ve found that using detection zones create misses or delays in recording events that I really want to see.

[Mod Edit]: Title Modified to Enhance Search Clarity.

The detection is triggered most of the time due to the shadow changes. There is no movement. Sometimes, it detects movement due to the wind and not vehicle/person moves. In outdoor setting, this is false triggering. A vehicle was parked in front of the camera, but the shadow or wind blowing causing some motion on other things, such as tree, bushes and the detection indicates vehicle. The car never moves. AI should checks to see if a vehicle is detected and see if that vehicle is moving or not. If not, then the detection should not mention about car. Also, when motion is detected due to wind or changes in brightness, there is no heat information and AI should not be triggered.


I"m finding that most detections on the camera focused on the front of my house are caused by the wind moving bushes/trees and shadows changing. Less than 5% of the alerts are relevant, less than 1% on windy days. This makes the feature pretty much useless.