Re:Tesla Earning Result 3Q 2020 [pdf]
by SEJeff on Wednesday October 21, 2020 @06:09PM
(#24852773)
The current system is a separate neural network running against every frame of every one of the 8 cameras. The new one is what elon calls the "4D" image processing. It uses all cameras to create a composite 3d representation where the 4th dimension is time. One of the things they use this for is to predict where an object will be in the future.
[ Parent
| on HN ]
There are no comments on this comment yet.