Bird's eye view coming?

[SO16]

A friend of mine has a $40K FORD with self parking and birds eye view, I was very impressed how well it worked in a tight reverse park. He did say to me 'your Tesla would have this right?' , I'm thinking yeah I've seen the Tesla self parking fails on YouTube, lets change the subject.

You can find videos and articles about Ford self parking failures too.

 

[P100D_Me]

Oh I'm sure I could, but his car is a $40K FORD, not a $140K self proclaimed technical marvel. After living with Tesla's terrible software QA and AP glitches for a while now I simply wouldn't trust the same computer systems to park my car.

 

[scaesare]

IMHO people who would rather park the car themselves given a perfectly reliable self-parking system are going to be very, very rare. Perhaps some parking skill competition, who can park more perfectly, but those will likely disallow usage of bird's eye view anyways. So, the use-case for birds eye view in a car which reliably parks itself as good as a human (or better according to Elon) is very niche.

Ah, so you admit it's non-zero. Good.

So, regardless of whether the usage take on vector-view is 5%, or 95%, there's some value in non FSD use cases. (And for the record, I disagree with your "vary, vary rare" guesstimate).

So now that we've established that, it logically follows that the introduction of such a feature does not preclude the successful implementation of FSD, as there are a subset of drivers/cases where vector-view would be used regardless.

Back to my earlier question, when do you think FSD will be complete? If you agree not for a while, it makes sense to have birds-eye-view as people will be parking manually for a very long time. If you think it's coming this year, birds eye view is not very useful. Note that even if regulators don't allow FSD, if Tesla has the software done, they could enable FSD self-parking as that is already allowed by regulators.

I don't really know. That's the problem with doing things that have never been done before. Prediction is very hard. Elon knows more about it than you or I do... but he's also been optimistic.

That having been said, I don't think vector-view introduction is any real indicator. If the work being done with vision allows V-V to be available essentially as a free by-product, and customers want it, then why not release it, regardless as to if FSD is 6 months or 6 years away?

 

[AudubonB]

I don't know how long the "look-at-the-screen-for-two-images-of-your-rear-sideviews" feature has been available; I discovered it just two days ago, but its existence suggests that Tesla could catch up to where other auto companies were 5-12 years ago with bird's-eye....¿soon?

 

[TravelFree]

First of all, I've had this on my Nissan Leaf and it is worth it for parking in tight spots. The showroom advisor told me the Tesla uses tilt mirrors to achieve the same thing. Not the same! Bird's eye view is great for parking lots.

As for FSD- I agree it has nothing to do with Full Self Drive. Summon is about as close to it as we have now. But each new release I find the new FSD or AP stuff unreliable loke stop sign recognition and green light go and only use FSD in the highway. But I have to initiate the passing lane changes or too often I miss my exit if traffic is heavy.

Anyway, Tesla should just add BEV to the basic package. Then change the FSD name to Navigate on Autopilot as that is what it's called in my Tesla.

 

[drtimhill]

The basic problem with a lot of these convos is it quickly divides into two camps:

1. Those who purchased the FSD package. They want new features like V-V to be FSD exclusive as it justifies (to them) the purchase price of FSD and gives them the feeling of feature exclusivity.

2. Those who did NOT purchase the FSD package. They want features like V-V to not require FSD so they get access to the new toys and fun features, without having to pay $$$ for FSD.

And that. imho, is really what drives the discussion. All the other minutiae of "does this need xxx" or "it needs xxx so must be in FSD" arguments are just rationalizations for one of the two positions above (imho, of course).

 

[whitex]

f the work being done with vision allows V-V to be available essentially as a free by-product, and customers want it, then why not release it, regardless as to if FSD is 6 months or 6 years away?

I agree Tesla needs birds eye view in order to compete, but only because FSD parking is not coming any time soon. I disagree that it's a free byproduct - DNN don't see the surroundings in the way birds-eye-parking-assist do, so it requires additional software development. Stitching all the video, and especially filling-in front view with manipulated old images (since there is no front bumper camera) will not just take software development, QA, and maintenance, but also compute cycles. Of course, given FSD doesn't do much of the originally advertised functions, maybe it doesn't use up most of the hardware so those cycles might actually be almost free (only cost is electricity since the hardware is already there doing not much). As for customers, it's a consolation prize for spending $8,000 on a feature which has been in vaporware status for the last 4 years.

Birds eye view really should be a much cheaper option, or even free on cars at Model S/X/Roadster price range. Obviously it has nothing to do with Full Self Driving functionality - it is clearly intended for Full Manual Driving. It is no different than if Elon was to tie new games to FSD package to make it feel more worth the money (heck, you could argue games can be used during autonomous driving, so more related to FSD than birds-eye-view-parking-assist).

 

[viper2ko]

I agree Tesla needs birds eye view in order to compete, but only because FSD parking is not coming any time soon. I disagree that it's a free byproduct - DNN don't see the surroundings in the way birds-eye-parking-assist do, so it requires additional software development. Stitching all the video, and especially filling-in front view with manipulated old images (since there is no front bumper camera) will not just take software development, QA, and maintenance, but also compute cycles. Of course, given FSD doesn't do much of the originally advertised functions, maybe it doesn't use up most of the hardware so those cycles might actually be almost free (only cost is electricity since the hardware is already there doing not much). As for customers, it's a consolation prize for spending $8,000 on a feature which has been in vaporware status for the last 4 years.

Birds eye view really should be a much cheaper option, or even free on cars at Model S/X/Roadster price range. Obviously it has nothing to do with Full Self Driving functionality - it is clearly intended for Full Manual Driving. It is no different than if Elon was to tie new games to FSD package to make it feel more worth the money (heck, you could argue games can be used during autonomous driving, so more related to FSD than birds-eye-view-parking-assist).

Odd that it will be MCU2 only but he is calling it a part of fsd

 

[scaesare]

I agree Tesla needs birds eye view in order to compete, but only because FSD parking is not coming any time soon. I disagree that it's a free byproduct - DNN don't see the surroundings in the way birds-eye-parking-assist do, so it requires additional software development. Stitching all the video, and especially filling-in front view with manipulated old images (since there is no front bumper camera) will not just take software development, QA, and maintenance, but also compute cycles.

Probably not literally "free" as in absolutely no coding required. But FSD development is going to require the FSD neural net to build a 360 degree model of it's surroundings in order to navigate through space. We've already seen that in action in the Karpathy videos of the car modeling the parking lot it's driving through in "birds-eye" vector-space (which incidentally I suspect is more like what this feature will be like).

So will there be a non-zero development effort to take that NN vector representation and build a pretty presentation on the MCU? Yes.

But my point is that the work for this to be possible is falling out of the FSD development, so releasing it as an additional separate feature is a minimal effort. And doing so in no way implies anything about FSD's future availability.

Of course, given FSD doesn't do much of the originally advertised functions, maybe it doesn't use up most of the hardware so those cycles might actually be almost free (only cost is electricity since the hardware is already there doing not much). As for customers, it's a consolation prize for spending $8,000 on a feature which has been in vaporware status for the last 4 years.

Birds eye view really should be a much cheaper option, or even free on cars at Model S/X/Roadster price range. Obviously it has nothing to do with Full Self Driving functionality - it is clearly intended for Full Manual Driving. It is no different than if Elon was to tie new games to FSD package to make it feel more worth the money (heck, you could argue games can be used during autonomous driving, so more related to FSD than birds-eye-view-parking-assist).

It seems you really want to use this topic as a venue to voice your dissatisfaction over FSD capability and schedule... so I've made my point, and won't belabor it any further.

 

[ItsNotAboutTheMoney]

Probably not literally "free" as in absolutely no coding required. But FSD development is going to require the FSD neural net to build a 360 degree model of it's surroundings in order to navigate through space. We've already seen that in action in the Karpathy videos of the car modelling the parking lot it's driving through in "birds-eye" vector-space (which incidentally I suspect is more like what this feature will be like).

So will there be a non-zero development effort to take that NN vector representation and build a pretty presentation on the MCU? Yes.

But my point is that the work for this to be possible is falling out of the FSD development, so releasing it as an additional separate feature is a minimal effort. And doing so in no way implies anything about FSD's future availability.

Other benefit is that any features added to FSD allow Tesla to claim some FSD money.

 

[BigNick]

FSD complete? Never, not in the lifetime of my car. Not in the level 4 or 5 sense, anyway.

My wife won't let me use autopilot when she's with me - which is 99% of the time. She's been scared too often, especially with phantom braking.

As for self parking - I've tried it some 20 times and it's worked once.

The software is so buggy it's a joke: only just now the sentrycam display was obscured in the bottom 2" by the menu bar, so I couldn't see the two bottom thumbnails or the progress bar. Had to reboot and I haven't checked it's changed yet. I could go on.....

My wife didn't like me using AutoPilot either. Her new car has (camera-only) adaptive cruise control which she likes to use, so she's not as scared of the Tesla AP anymore.

Her system has occasional phantom-braking and late-braking incidents, too, and its basic lane monitoring and "wheel-nudge" lane keeping assist doesn't qualify as auto-steer, though it will throw a "Hold the Wheel" type of warning if it detects you're not actively steering.

I still think AP2+ still has yet to reach AP1's level in some basic ways. But in many other ways, it has far surpassed AP1. And AP1 is still more advanced than the other guys I've tried (GM - although I haven't tried Cadillac's SuperCruise and Nissan, mostly.)

I think Tesla's "virtual rumble strips" when you stray from your lane are the most attention-getting alert. Better than just a beep or "alert seat", you hear it and feel it in the steering wheel and it feels nearly identical to the real rumble strips some roads have in their center lines.

 

[TravelFree]

I think Tesla's "virtual rumble strips" when you stray from your lane are the most attention-getting alert. Better than just a beep or "alert seat", you hear it and feel it in the steering wheel and it feels nearly identical to the real rumble strips some roads have in their center lines.

I never knew what to call it. My Leaf has that whenever you change lanes without the turn signal. The Tesla doesn't but if I get over the edge of the road on the right it activates loudly.