-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Terminator857
Active Member
Poor prediction of other cars seems to be the bottleneck. Elon has also aknowleged prediction as being a big focus for their next minor release, so maybe 9.2 will see some improvement in this specific scenario.
From picture we can see, that that truck is not in the visual field of the main forward camera or the narrow forward camera. Wide forward camera can only see 60 meters, so the truck is probably beyond its range.
According to Tesla forward looking side camera has 80 meters max distance.
What is the speed of the oncoming traffic in that crossing? If it is 50 mph (80 kph), forward looking side camera has 3.6 seconds window. I think that it is cutting close, whether FSD can do that turn safety. But I guess forward looking side camera should see that truck in that image?
Last edited:
Not showing on viz doesn’t mean it doesn’t see
I think the videos of V9.1 show nice improvements and forward progress, BUT I also think we still have a looonnnnggggggg way to go and a lot of little steps forward until wide release of full self driving. If I had to bet I'd say middle to late 2022 for FSD to go live with full capabilities while being safe.
We're closer than ever, but there's still a long way to go yet.
We're closer than ever, but there's still a long way to go yet.
Foresters have strange dreams
That would be awesome, seems like progress is being made. Since we are planting trees with a 60 year harvest target (and am never ever going to see that happen) I guess I am used to just focusing on progress..no matter how slowly.
Here we can see the obvious problem with forward looking side camera’s short range. To a human driver, fast approaching car in the middle lane is clearly visible as can be seen from the still image. But because forward looking side cameras have only 80 meters range, and wide forward camera 60 meters, they don’t see approaching car and FSD decides to go.
If car can keep itself angled so, that the oncoming traffic is in the field of the main forward camera, it might be ok.
Last edited:
powertoold
Active Member
I don't think we should draw any conclusions until Tesla releases more updates. We can spend time trying to figure out the exact distances or which cameras can see the cars, but it's just a waste of time at this point. The fact is that Chuck's wide angle camera can see the oncoming cars perfectly fine, so it's a matter of improving the NN / software.
My thought is that the vision-only NN is still dependent on the radar distance / velocity training data, so any cameras that weren't previously in the radar cone aren't good at predicting distance / velocity. It's unclear how Tesla will solve this, but they're definitely aware of it.
powertoold
Active Member
From the fact that Chuck’s camera can see the oncoming car we can’t necessarily draw the conclusion, that Tesla’s camera is capable to see it. With capability I mean optical capability combined with resolution. Software won’t help if the problem is in the cameras.
linux-works
Active Member
actually, that's what I'm afraid of; tesla saying 'ok' when its clearly not ready yet.
the hardware does not exist (that we know of) since its not functional enough to do level 3, let alone 4 or 5.
it would be irresponsible for them to release what they have now and let the unwashed masses have at it.
fastest way to get more restricive laws made, too. show enough bravado and release dangerous stuff early and we'll find out. tesla might actually end up ruining things for EVERYONE if they push too hard, too fast.
the fact that tesla will take your $10k does not mean that I want to be on the same roads as a bunch of yahoos running around with alpha-quality software controlling their several thousand pound killing machines.
Daniel in SD
(supervised)
Well it sure doesn't act like it sees it either! Also, even after the truck is displayed it flickers out of existence briefly. Why would it display the truck and then not display the truck?
I can see it quite clearly in a YouTube video. Isn't it close enough to the center that it would be seen by the long rang cameras? I'm sure it would be easy to see on the built in dash cam.
I don't really know anything about how they label all this video to train the NN. My understanding is that it's all done by hand so maybe they just haven't labeled cars more than 200ft away? I have no idea.
AlanSubie4Life
Efficiency Obsessed Member
“The FSD AI predictions are swiftly becoming superhuman.”
Just need to finish labeling.
“The improvement in 3 weeks is astonishing. Looks like we'll get wide release within 2 months at this rate of progress.”
Always be labeling.
powertoold
Active Member
Well ya, some aspects are becoming superhuman, not all aspects, yet.
AlanSubie4Life
Efficiency Obsessed Member
If perception still is not there at a level even approximating human vision, I am not sure what can really be superhuman at this point in time.
The current system works great at times to clean up certain human errors. A good use case for incomplete perception. That I guess is technically superhuman, but not what people are thinking in this context.
With a fully alert human performing the DDT, it is true that it is a superhuman system, although imperfect, with false alarms. That has been the case for several years though.
Similar threads
