Camera based Driver monitoring

[Microterf]

I'm going to make a prediction that in V11 (Holiday update) one of the things they roll out is driver monitoring with the camera.

Clearly they've been testing it based on the code, it will enable hands free for us and provide more assurance to regulators that Tesla cares about driver monitoring.

The obvious question is whether the cabin camera is up to the task, and what about S/X models.

 

[diplomat33]

I'm going to make a prediction that in V11 (Holiday update) one of the things they roll out is driver monitoring with the camera.

Clearly they've been testing it based on the code, it will enable hands free for us and provide more assurance to regulators that Tesla cares about driver monitoring.

The obvious question is whether the cabin camera is up to the task, and what about S/X models.

That would be amazing. I hope you are right.

 

[andrewFW]

@diplomat33 could post that he had eggs and toast for breakfast this morning and someone would click "disagree"

 

[mikes_fsd]

@diplomat33 could post that he had eggs and toast for breakfast this morning and someone would click "disagree"

@diplomat33 could post that he had eggs and toast for breakfast this morning and someone would click "disagree"

It would be terrible, because it would indicate one of 2 things at Tesla:

1. Elon does not believe true FSD is achievable with the current suite of hardware
2. Elon is out of the loop and someone was able to approve this project instead of working on making FSD better.

Also, why do I have to "agree" with his choice of breakfast food?

 

[andrewFW]

I think there are other potential choices there. The software may be lagging behind the hardware in terms of capability. Using the camera for driver monitoring could allow "hands free" before the software achieves the additional confidence needed for unsupervised hands free. I think it's a pragmatic choice.

 

[mikes_fsd]

The software may be lagging behind the hardware in terms of capability.

LMAO, so your solution is take engineering time away from solving that issue and throw them on an project that is DOA the second you achieve L4+ FSD?

This forum is full of comedians.

 

[arghx7]

LMAO, so your solution is take engineering time away from solving that issue and throw them on an project that is DOA the second you achieve L4+ FSD?

This forum is full of comedians.

You don't know what kind of discussions are going on with regulators behind the scenes. NHTSA may have requested this as part of a road map for approval of level 4 operation. Give them the benefit of the doubt.

 

[mikes_fsd]

You don't know what kind of discussions are going on with regulators behind the scenes. NHTSA may have requested this as part of a road map for approval of level 4 operation. Give them the benefit of the doubt.

You don't either, so why make up imaginary obstacles when there are plenty of issues to solve at hand??

 

[diplomat33]

It would be terrible, because it would indicate one of 2 things at Tesla:

1. Elon does not believe true FSD is achievable with the current suite of hardware
2. Elon is out of the loop and someone was able to approve this project instead of working on making FSD better.

LMAO, so your solution is take engineering time away from solving that issue and throw them on an project that is DOA the second you achieve L4+ FSD?

This forum is full of comedians.

Not doing a camera based driver attention system only makes sense if you believe Tesla really is super close to achieving driverless FSD. Do you really believe that Tesla is super close to driverless FSD where driver attention is not needed?

I don't believe that Tesla is super close to driverless FSD. So I think it makes sense to spend resources on adding a camera based driver monitoring system for several reasons:

1) It would allow hands-free immediately while we wait for driverless FSD.

2) It could also be used to add safety for people who don't have FSD.

3) It would allow Tesla to release FSD sooner, with driver supervision, and have a more reliable way of making sure the driver is monitoring FSD.

 

[helvio]

LMAO, so your solution is take engineering time away from solving that issue and throw them on an project that is DOA the second you achieve L4+ FSD?

This forum is full of comedians.

I'd say adding facial recognition is trivial. And adding the hardware would not ultimately be a waste of time, video conference is a possibility.
Mid game: use camera to monitor driver
End game: stop monitoring the driver and keep video conferencing capabilities (unavailable for S/X)

Not disagreeing, just stating a possibility!

 

[mikes_fsd]

Not doing a camera based driver attention system only makes sense if you believe Tesla really is super close to achieving driverless FSD. Do you really believe that Tesla is super close to driverless FSD where driver attention is not needed?

I don't believe that Tesla is super close to driverless FSD. So I think it makes sense to spend resources on adding a camera based driver monitoring system for several reasons:

1) It would allow hands-free immediately while we wait for driverless FSD.

2) It could also be used to add safety for people who don't have FSD.

3) It would allow Tesla to release FSD sooner, with driver supervision, and have a more reliable way of making sure the driver is monitoring FSD.

I do believe they are close.
Introducing driver monitoring would distract a team wholly focused on getting FSD working and fragmenting it to focus on sideshows.

I'd say adding facial recognition is trivial. And adding the hardware would not ultimately be a waste of time, video conference is a possibility.
Mid game: use camera to monitor driver
End game: stop monitoring the driver and keep video conferencing capabilities (unavailable for S/X)

1. you do not need facial recognition for conferencing.
2. The VAST majority of Tesla's on the road have cabin camera, the small percentage of Model S + X can plug in a USB camera if that was such a high demand item.
3. Monitoring driver software still would require engineering resources but would leave out the Model S and X crowd if it was to become available.
4. Then you still have your the usual problems of glasses (vision correcting and sunglasses).

 

[mikes_fsd]

https://twitter.com/elonmusk/status/1315448376163524608

 

[helvio]

I do believe they are close.

1. you do not need facial recognition for conferencing.
2. The VAST majority of Tesla's on the road have cabin camera, the small percentage of Model S + X can plug in a USB camera if that was such a high demand item.
3. Monitoring driver software still would require engineering resources but would leave out the Model S and X crowd if it was to become available.
4. Then you still have your the usual problems of glasses (vision correcting and sunglasses).

I see your points, but most of them have been addressed one way or another, IMHO.
1. Matter of fact, you do. Smart clipping is a pretty common feature nowadays for any cameras. Also, it's already trivial.
2. Completely agreed.
3. S/X can still use the wheel. It's not like you'll miss something you never had.
4. It's already addressed. I know you don't like green (I'm not in love with his snob either) but he notes that there's a few results that account for those situations. I just found the link:
green on Twitter: "In case you were wondering what does the selfie camera in model 3 currently try to detect: BLINDED DARK EYES_CLOSED EYES_DOWN EYES_NOMINAL EYES_UP HEAD_DOWN HEAD_TRUNC LOOKING_LEFT LOOKING_RIGHT PHONE_USE SUNGLASSES_EYES_LIKELY_NOMINAL SUNGLASSES_LIKELY_EYES_DOWN" / Twitter

 

[J1mbo]

It is starting to look like camera-based driver monitoring will become a regulatory requirement for self-driving, at least while there's still a steering wheel in the car.

If S/X are refreshed next quarter then it would be a good time to add an internal camera, and once the whole range supports it, Tesla can roll camera-based monitoring out without having some Heath Robinson solution for S/X. Current S/X owners with FSD obviously lose out, but it won't be the first time that's happened.

 

[arghx7]

You don't either, so why make up imaginary obstacles when there are plenty of issues to solve at hand??

Exactly how is regulatory acceptance and public perception of the technology an imaginary obstacle? When you have widely read publications like Consumer Reports attacking Autopilot for poor driver monitoring, it's important to put the public and the bureaucrats at ease. I deal with regulatory certification and diagnostic monitors of cars in my day job, and trust me, behind the scenes the Feds push for stuff all the time. You just never hear about it.

 

[Snerruc]

You don’t think people will cheat a camera? I guess you didn’t see the pics of glasses with eyes on them posted in another thread. Unfortunately, there are plenty of people who love to cheat things like this.

 

[helvio]

You don’t think people will cheat a camera? I guess you didn’t see the pics of glasses with eyes on them posted in another thread. Unfortunately, there are plenty of people who love to cheat things like this.

Oh my! Please send me a link to that!

 

[Microterf]

Of course there will be people that cheat it to show that it is cheatable. 6 months ago, I wouldn't have made this prediction because I didn't know they were working on it, but it appears there is already code in the system to start using the camera for driver monitoring.

Personally, I think it will lead to a smoother experience, especially for NoA with lane changes. I don't know about you guys, but I've found it's very smooth when I use a weight to defeat the nag, but when I don't have that, it's annoying as you have to nudge the wheel once, then right before it changes lanes, you have to nudge it again and for some reason it seems more sensitive then, and disengages AP quite often.

Just from a convenience perspective, I think it would be easier to just watch the road and let the car do its thing.

 

[diplomat33]

@diplomat33 could post that he had eggs and toast for breakfast this morning and someone would click "disagree"

Getting a "disagree" from @mspisars is a badge of honor. It means I posted something that is accurate and true.

 

[EVNow]

widely read publications like Consumer Reports

LOL. Fox News/CNN get just a million or so viewers a day (out of 100 million+ US adults). How widely is CR really circulated / read ?

Clearly they've been testing it based on the code, it will enable hands free for us and provide more assurance to regulators that Tesla cares about driver monitoring.

I don't think they will roll it out without having first tested using EAP.

But I agree, it is one of the things they could roll out - mostly to satisfy critics. But if it reduces the nags I get - I more than welcome the Big Brother. Good thing is - even if the camera is not very good at detecting correctly, they will only have to throw a nag, which they do anyway do now.