No disrespect to Bjorn, but that was a pretty flawed test.
Yep, but my main point was that I dont think its processing all that video real time like AP.
I think what it does is something simpler, like recording to dashcam and IF parking sensors are triggered, saving a 10 min clip to a different folder.
I do not think it's running full recognition of cars/people/cyclists.
Of course I could be wrong. Would be interesting test....
1 Park the car in a field with sentry mode.
2. Walk out of range of parking sensors and give it a good 10 min to forget you were there
3. Throw a water balloon (or something that won't damage the car) at the car roof (making sure not to trigger parking sensors)
4. Leave it for a good 10 min - don't walk up to it
5. Come back and see if it saved the footage or event
I bet that's how they do it. Running all that image processing all the time would probably be quite resource hungry.
Read a few reports where people get tons of events in their drive with no people... I'm betting that's some nearby bush triggering parking sensors at the range limit.