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

Tesla AI Day - 2021

This site may earn commission on affiliate links.
Yeah, at an initial glance, the highly detailed recreated point cloud of the world seems to have x-ray capabilities seeing through walls, but the data includes the whole video from all cameras including views from the future, which have different perspectives to see "through" and "behind" other objects.

View attachment 699111

Here's the intersection from Mexico City where the car is making a left turn, so the far side of the building on the right is only briefly visible from the right pillar and getting further away from the left repeater, yet it can recreate the entryways on that building.

It's incredible, borderline unbelievable, that information from cameras can generate such detailed point clouds on one pass. I know they're using views from the future and whatnot, but still...
 
what last night proved is after all the people who have been saying that simulation is basically a crutch and basically useless including Tesla. Are now parading around talking about Tesla “amazing” simulation.

I remember posting videos of the exact same thing Tesla was demoing and I was thumped down every single time and refuted. Shows you that some Tesla fans are god tier clueless even for things that benefit them. Unless they hear it from their gods mouths they will ridicule it. True definition of mindless fanaticism
 
I am curious if the "next gen cameras" coming with HW4 will just be better cameras or whether it will also mean additional cameras and/or different camera placement.
Andrej rectifying examples in the presentation was really interesting to take all the models (S/X/Y/3) different camera locations (height, dist from each other, etc) and convert them to the "same" video perspective seems relevant. I haven't seen/heard anything in there indicating where they are missing pieces of the 360 degree view.
 
Agreed the Dojo presentation was amazing. I think Tesla Bot is very much in keeping with Tesla's aesthetic, in that it goes for the sleekest form factor from day one (no protruding backpacks/wires/etc, just as the car has no LIDAR/antennas), and the tech follows within that constraint. The haiku method of tech development: impose the aesthetic constraints first, then see what you can do within that context. It seemed obvious to me that Tesla Bot will be pure blue-sky research and not an actual consumer or business product for a very long time; my guess would be at least a decade. It did invoke shades of "Klara and the Sun", a great book if you haven't read it. (On Obama's summer reading list.)

But will it have uneven panel gaps? :p
 
  • Funny
Reactions: DelPhonic1
Andrej rectifying examples in the presentation was really interesting to take all the models (S/X/Y/3) different camera locations (height, dist from each other, etc) and convert them to the "same" video perspective seems relevant. I haven't seen/heard anything in there indicating where they are missing pieces of the 360 degree view.

Well, one reason I asked the question, is because we've seen FSD Beta videos where it would seem that additional cameras could be helpful in certain cases. So I think it will be interesting to see if Tesla does add extra cameras or not.
 
I'm likely missing something obvious, but I wonder about all the NN's focus on the short time horizon for an individual car. Isn't there a great opportunity for at least each Tesla car to share it's surrounding 3D 'image' to nearby Teslas?

For example, one Tesla could tell an oncoming Tesla "Watch for and avoid the child who's fallen off their bicycle at location x + y".
its called v2x (vehicle to anything) and I'm 100% a supporter of that kind of tech. I think it will HAVE to be part of the safety system that will make level 5 actually work. I'll say this, even: level 5 wont have enough nines until that closed feedback loop is, well, closed. having something other than the car give you a reality check, that's a must-have in my book.

watch v2x and see if vendors embrace it. maybe other countries will try it first. it really is a key tech, for so many reasons.
 
What do you all think about the things Karpathy said about large vehicles / trucks not fitting into a single camera so they did camera fusion to locate. Does that give us insight into why Tesla has the numbers of cameras they do and at the locations they have selected? It seems on lots of other threads, people always start wondering why we don't have 4 cameras at the corners of the cars. I'm sure the number & placement is not an oversight on Tesla's part. Did any of you glean any tidbits around this?
 
  • Like
Reactions: scottf200
Well, one reason I asked the question, is because we've seen FSD Beta videos where it would seem that additional cameras could be helpful in certain cases. So I think it will be interesting to see if Tesla does add extra cameras or not.
The only obvious one I've seen is in Chuck's video when there is occlusion to the left | right at an intersection and you are first in line (bush or fence or utility box). That would seem to be helped with left | right cameras near the headlights. The current solution is to 'creep' out like a human would do since they can't see either.
 
  • Like
Reactions: Dan D.
What do you all think about the things Karpathy said about large vehicles / trucks not fitting into a single camera so they did camera fusion to locate. Does that give us insight into why Tesla has the numbers of cameras they do and at the locations they have selected? It seems on lots of other threads, people always start wondering why we don't have 4 cameras at the corners of the cars. I'm sure the number & placement is not an oversight on Tesla's part. Did any of you glean any tidbits around this?
That was a great section of the presentation. The example below with larger or occluded vehicles.

Vst5UC2.jpg
 
  • Informative
Reactions: Dan D.
What do you all think about the things Karpathy said about large vehicles / trucks not fitting into a single camera so they did camera fusion to locate. Does that give us insight into why Tesla has the numbers of cameras they do and at the locations they have selected? It seems on lots of other threads, people always start wondering why we don't have 4 cameras at the corners of the cars. I'm sure the number & placement is not an oversight on Tesla's part. Did any of you glean any tidbits around this?

I don't think the camera position is directly related to large vehicles not fitting into a single camera. Tesla's camera position is probably to cover relevant angles while driving. Specifically, you will note that Tesla's cameras seem to correlate to human driving. There are front cameras as humans have eyes that look to the front. There is a rear camera that is equivalent to looking in the rear view mirror. There are side repeater cameras that are equivalent to looking in the side mirrors. And the B pillar cameras are somewhat equivalent to when the human turns their head to look sideways at intersections. They are positioned a bit further back to the human driver but the B pillar is the natural location for the cameras.

The only obvious one I've seen is in Chuck's video when there is occlusion to the left | right at an intersection and you are first in line (bush or fence or utility box). That would seem to be helped with left | right cameras near the headlights. The current solution is to 'creep' out like a human would do since they can't see either.

Yeah, that is one scenario. And yes, the car can creep forward but creeping forward, especially with busy cross traffic, is not ideal. Adding cameras would allow the car to see with less forward creeping and therefore could add safety.
 
  • Like
Reactions: Dan D.
What do you all think about the things Karpathy said about large vehicles / trucks not fitting into a single camera so they did camera fusion to locate. Does that give us insight into why Tesla has the numbers of cameras they do and at the locations they have selected? It seems on lots of other threads, people always start wondering why we don't have 4 cameras at the corners of the cars. I'm sure the number & placement is not an oversight on Tesla's part. Did any of you glean any tidbits around this?
I'm certain that the 8 cameras were placed there as a best guess at the time as to what might be needed. They did give Tesla 360 coverage with some overlap. However as the system has evolved they turn out to have large blind spots depending on the position of occluding objects. Also there are blind spots close to the car at low heights. This hampers FSD as well as birds eye view and self-parking.

Their main problem is a stubbornness to admit this. If they gave up and created a new camera system I'll bet they would move ahead very rapidly. It would mean a huge problem to their existing cars on the road. That is what is wrong with their approach IMO. It's sad that they are so advanced in methods yet so stuck with the past hardware.
 
Your comment is confusing because this was *very* specifically talked about in the presentation. This was in a few places but here is one:

hQgUp00.jpg

Let me clarify. I meant I don't think Tesla picked 8 cameras and their locations in order to view large vehicles that span multiple cameras. They picked 8 cameras and their locations to give the car a suitable 360 degree view around the car. But Tesla was faced with the problem of different cameras seeing the same large object and thinking they were two different objects. AI Day talked about how Tesla solved that problem.