Whilst this obviously seems like a dumb move (and likely will stay that way) is there not a case that the vision software will continue improving rapidly until it is better than the original parking sensors?
IMO the major benefit of parking sensors is that they see things that the driver and the cameras can’t. Not because of perceptual or image processing issues — because the obstacles are obscured by the rest of the car.
So surely a good enough machine vision algorithm could do the same things if it had camera on the bumper, but they would get dirty and be more expensive than ultrasonic sensors on the bumpers.
How many times have you not been sure whether there’s an obstacle right in front of your car because you can’t see there?
(Maybe a really nice mapping algorithm could remember what’s in the way, but that’s not terribly useful when un-parking if obstacles move, or when parking and approaching obstacles of which the driver and the cameras have never gotten a good view.)
I'd just like functional top-down 360 camera view like a bargain basement Nissan Rogue has. My eyes can detect the curb.
Part of the problem is that the cameras Tesla puts in a Model 3/Y were mediocre in 2018. By today's standards they're really quite bad. Though I think I heard a rumor that this would finally be changing soon.