Hi,
When I learned Tesla went from frame-pixel analysis to object classification, I knew it would take awhile to get a better product. In effect the earlier approach was suspended.
Pixel analysis does not scale well and has no ability to predict paths. In contrast, classification of objects means they follow predictable paths. The computational load of a set of objects is trivial compared to analyzing the pixels of each frame and throwing it away with the next frame.
FSD is still 'beta' which means there are reproducible errors. Happily the beta testers are pretty good about sharing the weaker parts.
Being resentful about the development delay does no good. As a beta tester, I'm impressed with the current performance, a C+ or B- grade.
Bob Wilson
__________________
2019 Tesla Model 3 Std. Range Plus - 215 mi EV
2017 BMW i3-REx - 106 mi EV, 88 mi mid-grade
Retired engineer, Huntsville, AL
Last edited by bwilson4web; 02-08-2022 at 03:36 PM..
|