eswimm
Well-known member
- Joined
- Jun 6, 2024
- Threads
- 3
- Messages
- 555
- Reaction score
- 994
- Location
- Charlotte, NC
- Vehicles
- Model Y, Cybertruck
Part of the issue is that FSD needs to be trained for everything you expect it to recognize. Right now, it's unlikely that FSD knows the difference between a chicken and an empty plastic bag blowing across the road. Once you throw it training data for a menagerie of animals, then you need to determine how it's going to react. Maybe you decide to swerve for a person even if you risk losing control, but do you swerve for an animal? Normal thought is you brake for an animal if you can, but it's usually better to hit it than to swerve.
The more objects you define, the more complex the identification becomes, how confident is FSD that it recognizes any given portion of the image as an object. It's the reason you sometimes see phantom people and semis when you pull into your garage, FSD starts seeing people and vehicles in the various objects in frame.
The more objects you define, the more complex the identification becomes, how confident is FSD that it recognizes any given portion of the image as an object. It's the reason you sometimes see phantom people and semis when you pull into your garage, FSD starts seeing people and vehicles in the various objects in frame.
Sponsored