False Positives
Hi Ben;
I'm experiencing false positives caused by a tree in the front yard. Photo attached below for reference.
On viewing the motion detection in the preferences window, I'm seeing that the software is picking up movement in the form of stray branches that go beyond the greened off mask I have put in, and also on the shadows cast on the sidewalk. This movement is being picked up, triggers the AI, which then seems to see a parked car on the street, registering it with '100%' confidence and sending off a notification even though it's not the item that's moving.
Is there any way to avoid this beyond turning off vehicle notifications or masking the entire street? I rely on the notifications to let me know when deliveries have arrived, so I would prefer not to disable vehicle detection if I didn't have to.
Comments
Cut the tree
SecuritySpy's vehicle AI is very accurate (>98%) at determining whether there's a vehicle in the image, however it cannot (currently) tell the difference between a moving vehicle and a stationary one. Therefore, when you have a view where there is often a static/parked vehicle, you basically can't use the vehicle AI, because it will give you exactly the outcome that you are experiencing, whereby other movement such as shadows and tree branches cause motion detection around the vehicle, and then the AI detects the vehicle, causing a trigger.
You have two ways of dealing with this. The first is to change the camera's view so that static vehicles aren't in the frame, but only moving vehicles could be (e.g. a delivery vehicle entering your driveway). Alternatively, turn off vehicle detection and use human detection only - this will catch delivery personnel leaving their vehicle and walking to your house. And, with human detection, there is no similar problem of "parked" stationary humans to deal with!
Above, you say that it cannot "currently" detect whether a vehicle is moving. Is it possible AI will eventually detect motion as well as human, vehicle, or animal? And if so, any rough estimate of when it might happen?
I'm asking because I have a similar issue (as well as the exact same one), we have a statue on our front porch which AI detects as either human or animal whenever there is motion in the frame. This happens every sunny afternoon when the shadows of moving leaves fall onto the porch.
This makes it a little annoying to have to go through the many motion generated recordings to see if there actually is a person or animal on the front porch.