Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Are FSD visualization pre-rendered/pre-made?

This site may earn commission on affiliate links.
I'm trialing V12.3.4 FSD right now and come to realize that the FSD visualization is just too good to be true to be done in real-time by the cameras on the vehicle. Here are a list of things that made me think that the visualization is a hybrid of what the cameras see and what's been premade based on crowd sourcing or map data:
  1. When I'm on an overpass and it would partially render a road that was below me or another overpass that's higher than me. The camera can't see them at all.
  2. Render the opposite lanes perfectly even though there's a 3ft tall concrete road barrier in between.
  3. Driving behind a bus and it render the roads ahead perfectly.
  4. In a busy intersection it could render all the roads perfectly even though most of the sightlines are blocked by cars.
I think the reason the basic Autopilot is not getting the detail FSD visualization is that it's actually generating what the cameras see without the premade visualization. Free users are not getting the Premium HD Maps that was generated for FSD.
 
the FSD visualization is just too good to be true to be done in real-time by the cameras on the vehicle.
The car has access to road data, so it can do a lot of stuff with the road topology based on knowing where it is on the planet. We see that when cars are driving on snow-covered roads; the car still knows where the lanes are. But all the moving stuff is being identified and position in real time. I believe the lane lines are as well. Pretty much anything that flickers in and out of existence in the visualization is being perceived in real time.
 
The car has access to road data, so it can do a lot of stuff with the road topology based on knowing where it is on the planet. We see that when cars are driving on snow-covered roads; the car still knows where the lanes are. But all the moving stuff is being identified and position in real time. I believe the lane lines are as well. Pretty much anything that flickers in and out of existence in the visualization is being perceived in real time.
A neighborhood four-lane street plus shared center turn lane had recently received an asphalt maintenance coating and lanes hadn’t been repainted yet. The car on FSD estimated the lines — including the shared center turn lane - and seemed to be in the right place. An example where FSD was better than a human as I would have at best guessed on the four driving lanes but wouldn’t have attempted to guess where the shared, center turn lane was.
 
  • Informative
Reactions: JB47394