Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
If Tesla does change more than the DrivePX2, they will be in a hot mess. Rear facing radar, if needed, will be costly to install (I should think unless its somehow already wired). At any rate, they've produced like 80k HW2 vehicles and that's a massive retrofit (which I believe they would have to do for all vehicles when/if they opt for FSD).

I think it depends on "when" they introduce it. If they decide today oh shoot we need a rear radar to make FSD a reality, they could simply upgrade those that paid for full FSD which i'm betting is not that many via a SC appointment. And then when folks buy the FSD package later they can increase the price accordingly for the HW necessary and do it like the ludicrous upgrade, where it required a SC visit to enable. But if they wait until there's 200 - 300k FSD cars on the road before deciding, things get ugly quickly. As long as the FSD cars are very few, it should be a manageable item. However, once Tesla says we have FSD rolling out today via software update, everything hits the fan if it also requires a hardware upgrade.

That being said, I think Tesla has always know a PX2 upgrade is needed and they were just waiting for it to be ready, though I think they might skip a PX2 upgrade and just do Xavier for it's benefits in terms of cpu power and power usage, and skip the entire dual PX2 board thing, since it appears DOA from NVidia as of next year anyways in favor of the xavier platform.
 
If Tesla does change more than the DrivePX2, they will be in a hot mess. Rear facing radar, if needed, will be costly to install (I should think unless its somehow already wired). At any rate, they've produced like 80k HW2 vehicles and that's a massive retrofit (which I believe they would have to do for all vehicles when/if they opt for FSD).
For the life of me I can't see how a rear facing radar would be necessary under any circumstances... if anything front corner facing radar units would be more useful.
 
VeryGreen or it might be Wk057 have stated in another thread, they see evidence of a rear facing radar. Apologies, I forgot who said it, but someone with root access and bad-ss skills has said they have seen it. I'll see if I can dig it back up again, but yah there are hints of it.
Are you referring to this post? If so, it talks about a backup camera connection to AP, not about rear radar.
the way cameras are connected is different. Backup camera is now directly controlled by autopilot.
I am sure there are other changes, but there's only a very limited view into this at this time.

A rear radar install is going to be a much bigger deal than an ECU swap, and I can see how this would stretch the interpretation of the car having the necessary "hardware".
 
I'm surprised you can get images that look like this from an RCCC sensor. All you have is one red channel and the rest are monochrome.

As an example you're getting Yellow which is understandable since it's only Red and Green, but you're also getting shades of green.

@stopcrazypp already explained it very well and I have not much new to add. There is no Green, no Blue, not even Yellow. This is the whole palette used in my color visualizations:
palette.png


When I was seeing the first results in my experiments I could swore I can see blue sky and green information sign backgrounds, although I knew it was impossible. I'm not an expert in this field but I know our eyes see color quite poorly, and our brain is performing heavy "post-processing" to fill in missing information and make full picture in our head. There are optical illusion examples online (like this one) showing some strange properties of our sight (also, there is a slight chance some of you might heard about particular dress...). This is probably a manifestation of similar property of our brain.

It does all make sense, but I've just never imagined what an image would look like out of an RCCC camera. I've only used bayer RGGB cameras.

I was in exactly the same situation before this thread was created and I started experimenting to find it out. I don't know how people usually visualize the image from this kind of sensor.

I am familiar with sensors like this one where the electronics are all the same for the Mono/RGB/etc version, but the mask is different. Those frustrate me a bit since the gain of one channel is never quite the same gain as another channel even if set to the same value.

Yes, in my job we also found this problem with some sensors. It can also be seen in Autopilot sensor. Even in the image from the grayscale channel you can still see a checkerboard pattern. Theoretically it should be possible to correct this with some sort of flat field correction, but I never tested it. All my images was smoothed a bit with Gauss filter to make it less visible ("-ns" switch in the tool can be used to turn it off).
 
Are you referring to this post? If so, it talks about a backup camera connection to AP, not about rear radar.


A rear radar install is going to be a much bigger deal than an ECU swap, and I can see how this would stretch the interpretation of the car having the necessary "hardware".

Not that one, no this was specifically a rear radar when I get off of work I'll dig around and find it. Might have been on another site.
 
Here's what I think might be something Tesla might do regarding the AP2 suite: they'll update the computer and sensors (and call it 2.5 or 3.0) some time this fall on new cars. Then they'll say the new suite gets to FSD faster (because faster computer, better/more sensors) but the "old" 2.0 will get there too once the software matures/is optimized.

What do you guys think?
 
Here's what I think might be something Tesla might do regarding the AP2 suite: they'll update the computer and sensors (and call it 2.5 or 3.0) some time this fall on new cars. Then they'll say the new suite gets to FSD faster (because faster computer, better/more sensors) but the "old" 2.0 will get there too once the software matures/is optimized.

What do you guys think?
So far the code is mostly shared, as such splitting dev efforts like that sounds counterproductive from my view.
 
Here's what I think might be something Tesla might do regarding the AP2 suite: they'll update the computer and sensors (and call it 2.5 or 3.0) some time this fall on new cars. Then they'll say the new suite gets to FSD faster (because faster computer, better/more sensors) but the "old" 2.0 will get there too once the software matures/is optimized.

What do you guys think?

A big part of why I'm following this thread really carefully is I don't firmly believe the camera sensors will work very well in the rain. What evidence do I have of this? The rear camera seems rather bad in the rain. It might be a little better in HW2, but I haven't seen any proof of the dramatic improvement needed.

I think the camera performance in mild/bad weather will be a pretty major obstacle. But, hopefully I'm wrong.
 
A big part of why I'm following this thread really carefully is I don't firmly believe the camera sensors will work very well in the rain. What evidence do I have of this? The rear camera seems rather bad in the rain. It might be a little better in HW2, but I haven't seen any proof of the dramatic improvement needed.

I think the camera performance in mild/bad weather will be a pretty major obstacle. But, hopefully I'm wrong.
weather.png


That said, I covered backup camera with Rain-X and it has been a massive help.
 
A big part of why I'm following this thread really carefully is I don't firmly believe the camera sensors will work very well in the rain. What evidence do I have of this? The rear camera seems rather bad in the rain. It might be a little better in HW2, but I haven't seen any proof of the dramatic improvement needed.

I think the camera performance in mild/bad weather will be a pretty major obstacle. But, hopefully I'm wrong.
Besides cameras that face the rear, none of the forward facing cameras can get water directly on the lenses. The rear facing cameras don't really have a heavy workload. It's not like it's going to read the back of signs you've already passed
 
Besides cameras that face the rear, none of the forward facing cameras can get water directly on the lenses. The rear facing cameras don't really have a heavy workload. It's not like it's going to read the back of signs you've already passed

My concern is the rear view side spot monitoring cameras because my HW1 car is horrible at side monitoring without these. I haven't heard that the new ultrasonics are that much better.

I'm also concerned about the center rear camera for backing up into a parking spot. Seems kinda silly to have a snake charger, but then have the car ask me to clean the rear camera.
 
may I ask how do u turn the .h265 file to .mp4?
H.265 is an mp4... but you can convert using ffmpeg or handbrake.fr

My concern is the rear view side spot monitoring cameras because my HW1 car is horrible at side monitoring without these. I haven't heard that the new ultrasonics are that much better.

I'm also concerned about the center rear camera for backing up into a parking spot. Seems kinda silly to have a snake charger, but then have the car ask me to clean the rear camera.

The new ultrasonics are supposed to have greater range and currently there's no evidence AP2 uses the rear facing cameras at all, but if the goal later is detecting object vs non-object this should still be straightforward even with a few drops on the lens. The lens itself is fairly large:
Fw6pXAP.jpg


For it to fail in the future you'd have to have a set of circumstances where you are attempting to change lanes from a very slow speed into a lane of fast moving traffic coming from behind on the side where the camera is obscured. Generally in those circumstances, if your ultrasonics are clear enough to switch lanes, if you get rear ended then it's the fault of the driver who hit you. It remains to be seen how the car would actually react to an obscured view when they start using those cameras. Would it attempt to change lanes at all?

At the end of the day if the car is occupied, in the near future, it's the sole responsibility of the driver to double check the mirrors.
 
My rearview BUC stays mostly ok. Chicago has been nonstop rain for over 14 hours now and AP2 did just fine on the highway. It wasn't as rock solid as during clear weather but this was an exceptionally hard rain. Everyone else though was a disaster. One guy was going 25 in the fast lane without hazards on.

I did not push the system past 64mph as I was blinded at that speed and I'm not sure how the system operated except radar was working great. I kept a death grip on my wheel and disengaged AP twice over 12 miles due to flooded roads but it stayed in the lane (as far as I could see) just fine. It did wobble a lot more (less confident) and it did not present AP very often on local roads (which I didn't use because AP was unsteady on the highway). I used TACC on local roads and it worked great throughout (only once in 7 months has my radar gone out due to rain and that was because of intense spray from a semi for 3-4 minutes, after 10 minutes of being out of that spray it came back online).

I believe the cameras are sufficient in rain and snow. I haven't really tested it in heavy fog or a dust storm or something like that since I don't drive in heavy fog (we get it often in the northshore IL) and we don't have dust storms.