How to fix false positives
Comments
-
An animal (cat) has still been recorded despite the settings which should prevent it from being recorded.
Uploaded file is a .gif file.
From previous observations, I noticed that an animal would not have been recorded at daytime.
-
Hi @rapyt the issue here is that you have enabled the vehicle detector, but there is a stationary vehicle in the camera's view. The issue is that the AI is very accurate at identifying vehicles, but it cannot tell if the vehicle is moving or stationary. In this situation, if there is any movement around the vehicle (like this cat), the AI will see the vehicle and trigger recording. In this situation, you must either turn off the vehicle detector and instead just use the human detector, or you must draw a motion mask that completely covers any stationary vehicle (however I would not advise this in your case, as the stationary vehicles could move and so you would have to mask out a large area, which would reduce the chance of human detection in these areas).
-
@Ben, a while back in this thread I think you commented that one should mask out parked vehicles to reduce inappropriate vehicle identification.
How about considering 'masks per AI recognition net' => so a mask for each category.
This would be pretty useful; we park our cars in our driveway, and given household dynamics ( kids ) the parking isn't really predictable WRT a mask area. However we do want to know about vehicles passing by the end of our driveway, or in the front of the house.
Conversely when the driveway is empty we do infact want image recognition where the cars would usually be. In the scenario I am hinting that would be a 'recognize humans in this area' mask.
