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

2023 Model Y no USS discussion, incl. Discussion of Tesla Vision Firmware

This site may earn commission on affiliate links.
I had very high hopes/expectations from the update especially after so many months but very disappointed with it. It is not able to correctly detect distances to objects in the front and prompts to stop with almost 1 feet gap in the front when pulling into my garage. Also the animation of the lines on the front and sides are all wobbly and is not like how they advertise. Not sure if Vision can really replace USS/Radar.
To be fair, i dont expect Tesla to push this update and then abandon development.

This is V1.0 of the Camera based "USS"

I would kind of expect them to push this update and have it report back from the cars, so that they have massive amounts of data to use for further development and fine tuning of the system.

Some have reported that even with room to spare towards a wall, it would show that the car was already at the wall or "in" the wall. But this being V1.0 they are probably going to be cautios, so next version will probably be gradually better and they will move the limits a bit closer to what is the "truth".

Rather do this than not and have people complain they ran into a wall because the car reported that there were plenty of room.
 
To be fair, i dont expect Tesla to push this update and then abandon development.

This is V1.0 of the Camera based "USS"

I would kind of expect them to push this update and have it report back from the cars, so that they have massive amounts of data to use for further development and fine tuning of the system.

Some have reported that even with room to spare towards a wall, it would show that the car was already at the wall or "in" the wall. But this being V1.0 they are probably going to be cautios, so next version will probably be gradually better and they will move the limits a bit closer to what is the "truth".

Rather do this than not and have people complain they ran into a wall because the car reported that there were plenty of room.
Why doesn’t Tesla test it out first and achieve a high level of accuracy before they release something like this? Maybe before removing USS sensors. They can do this in-house in a controlled environment, unlike self driving.
 
  • Like
Reactions: glovelace
Yes, I got the 2023.6.9 with parking assist on my Model Y with USS removed. So far I am satisfied with
the new proximity function.
Hi all Pros out there. My 2023 MYLR got updated with 2023.6.9 yesterday. Not sure what that means yet, Please share with us your learning about this NEW proximity function. What does it do and why do U like it
Enjoy yours
 
  • Like
Reactions: glovelace
Why doesn’t Tesla test it out first and achieve a high level of accuracy before they release something like this? Maybe before removing USS sensors. They can do this in-house in a controlled environment, unlike self driving.
Because removing USS was a knee jerk reaction by Musk due to a combination of

1) Component shortages and
2) A newfound opportunity to cut costs

So once the word came down from Musk, the engineers were then tasked with solving the problem.
dennis-miller-my-opinion.gif
 
Loving the resounding yeehas on what people are seeing! The most common comment…I’m sure it’ll get better over time - hardly glowing endorsement…more muted thank goodness they have actually delivered something, even if it doesn’t work that well in lots of situations.
I've watched 2 Utubes related to 2023.6.9. Granted the feature is not perfect, not even close to perfection but at least it's doing something in the direction that I like to see. I would give it time for more iterations then the feature would become better. Remember how long it took Tesla to improve its manufacturing technique. Less complaints now , right?
 
  • Like
Reactions: WyoDude
Why doesn’t Tesla test it out first and achieve a high level of accuracy before they release something like this? Maybe before removing USS sensors. They can do this in-house in a controlled environment, unlike self driving.
Look i totally agree that they should have never removed the actual sensors before this was fixed, but as for the "Why did they not achieve a high level of accuracy" well, it will be totally impossible to test every single use case, every single obstacle at every single angle and so on, so its better to err on the side of caution.

Yes, this essentially means that the users will be "beta testers" but again....I would rather that it was a bit cautious in the beginning and then in a month or 2, they can dial the "safety margin" down a bit.
 
Loving the resounding yeehas on what people are seeing! The most common comment…I’m sure it’ll get better over time - hardly glowing endorsement…more muted thank goodness they have actually delivered something, even if it doesn’t work that well in lots of situations.
Well the non-USS cars have gotten better over time.

They went from "Does not have this feature" to "Have something that works somewhat good, most of the time"
 
I just got the update and I’m happy with it. I’d give it a 7/10. Having come from a M3 with USS to a vision only MY it’s nice to have the parking assist back. It’s definitely not as accurate as the USS but it’s better than nothing. I’ve noticed it’s a lot more conservative than the USS but better that than the other way around.

Honestly I was skeptical that we’d ever get parking assist back so having it this quickly has me pleasantly surprised.
 
I just got the update and I’m happy with it. I’d give it a 7/10. Having come from a M3 with USS to a vision only MY it’s nice to have the parking assist back. It’s definitely not as accurate as the USS but it’s better than nothing. I’ve noticed it’s a lot more conservative than the USS but better that than the other way around.

Honestly I was skeptical that we’d ever get parking assist back so having it this quickly has me pleasantly surprised.
" this quickly " : close to 6 months for some people. I've been without park assisy for 4 months now but EU based so no rollout yet here. Our definition of quickly does not match.
 
  • Like
Reactions: Flemingo
" this quickly " : close to 6 months for some people. I've been without park assisy for 4 months now but EU based so no rollout yet here. Our definition of quickly does not match.
Well it was quicker than my expectations, which were low. I realize that's a dumb thing to think for a $60K+ car, but such is life.

Overall I think the whole thing is stupid. The minivan I bought back in 2017 brand new for 27K has rear parking sensors and can even detect and warn you when a vehicle is approaching from either side while you're backing up. Yet here we have none of that.

So yeah, not trying to apologize for Tesla, cause I think they really screwed up on this one, just saying that I'm glad to be back to at least 7/10ths of where we used to be.
 
  • Like
Reactions: jeewee3000
I've watched 2 Utubes related to 2023.6.9. Granted the feature is not perfect, not even close to perfection but at least it's doing something in the direction that I like to see. I would give it time for more iterations then the feature would become better. Remember how long it took Tesla to improve its manufacturing technique. Less complaints now , right?
Maybe they should not have taken our money for features that are disabled (parking sensors being only one of the capabilities they withdrew). Maybe they should have considered customers a little more. Fraud is fraud. They did not deliver the car that I ordered and paid for. Do I want to be a beta tester for a capability that was there when I ordered and paid for my car, and actually worked better? No. Remembering how long it takes Tesla to deliver features is little comfort sadly. Personally no…no less complaints.
 
  • Like
Reactions: Mike7564
Not so accurate if it's a low-lying object, but still better than nothing. And that would make sense because how the hell is the camera going to be able to accurately predict how far away it is when it's out of sight?
This obstacle is well within my field of view from the driver's seat, so the front camera should easily be able to see it. Tesla Vision says it's 32 inches away, in reality it's less than 8 inches, I would argue that's actually worse than nothing. I agree though that this should improve over time. I hope the limited rollout is indicative that the development team is collecting a lot of data from situations like this and making improvements.
20230325_164029.jpg

...it takes forever to initialize when I first start out which renders it largely worthless with normal usage in and around parking lots and such. I'm sure this too will get better with time but it's the biggest initial gripe I have as not being available at all is worse than being a little inaccurate.
Same here; It has said "Park Assist Unavailable" in more than half of the situations where it would be useful, and that's excluding all of the times that it was unavailable due to water drops or road mist on the cameras.
How does it work for the back if the camera is covered with rain/fog?
Usually a drop of rain will render the camera useless. As for fog, I was thinking a mist over the camera which would distort any images (e.g., side cameras from condensation).
You are correct, park assist is completed unusable without clean, dry cameras. If you park outside, the current version is not useable in the mornings or evenings because of dew. If you get a drop of rain, road mist, or condensed fog on the lenses, they will have to be dried off in order for it to work.
They went from "Does not have this feature" to "Have something that works somewhat good, most of the time"
So far in my limited anecdotal experience, it's more like "Have something that may or may not work, in limited situations"
  • Situations where it works nicely:
    • Straight, well marked curbs (Yellow or red are best, it struggles more with curbs that are the same color as the pavement.
    • Parking blocks, especially if they are a different color than the pavement (concrete block on asphalt surface).
    • Parked cars
    • Walls and buildings
    • Even, indirect lighting, such as a cloudy (but not rainy) day at noon.
  • Situations where it struggles:
    • The first 10-30 seconds while it is loading after shifting out of Park.
    • Any obstruction or distortion of the cameras
      • Wet weather
      • Morning dew
      • Condensation
      • Fog
      • Road mist
    • Irregular obstacles
      • Landscaping, tree branches, shrubbery
      • signposts or other narrow objects
      • Curved curbs or corners of curbs
      • parked motorcycles
      • Bike racks with bicycles
      • Most non-car obstacles
    • Irregular lighting
      • Light source in same direction as obstacle (ie morning sun, evening sun)
      • Very bright, sunny situations, potentially causing glare.
      • Uneven lighting, mix of shadows and excessively bright areas
 
Park Assist upgrade came with this update (no FSD or EAP). Overestimates distance by about 40-50% in some cases and underestimates in some
5F6B1F1A-58F7-4506-B0D3-0AFE4B0195F0.jpeg
image.jpg
Keeps earlier position in memory and beeps the crap out if you parked too close to a barrier. That might turn some people off initially. A reasonable implementation and a good start. I think this will get better. (Have some experience in designing cameras and implementing imaging algorithms).
 
  • Like
Reactions: Sambub