Scenario B: The camera can’t directly see the bike rack, but can see the doorway you have to walk through to get to the bike rack. So somewhere in 4 hours of doorway footage, someone walks through the door, then a short time later walks back through the door with the bike. A binary search won’t help here
I never said it works 100% of the time. This that it would work most of the time. And I make that statement based on the fact that usually the environment changes around the event, or the event happens long enough to be detectable, if not by humans, then by AI.
In all of my comments I’m assuming that that focal point of the crime is visible.
But even if it wasn’t, if the person stealing the bike knocks over a trash can while doing it and that’s in the camera view it would still be useable. Or if a crowd congregates around the focus point and looks around for the bike, that would also make a binary search feasible.
That’s always just been my point, that a binary surgery more often than not works because most times the environment around the event changes in some way, from subtle to extreme.