-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Does FSD Beta use the steering wheel or the interior camera to know if youre paying attention?
- Thread starter EVchiq
- Start date
Tam
Well-Known Member
Yes. Both. If you think your current Autopilot nagging system is bothersome, it would get even stricter in FSD beta.
Definitely both. Sometimes I'll be paying attention to the road and it'll nag me about holding the steering wheel - and it'll turn out I was holding the wheel with both hands and balancing out the torque it needs for hand detection.
Picking up cellphone doesn't set off the red screaming wheel in my 2018 3 but it does give the usual nag.
There seems to be a lot of discrepancy between what gets reported here, and what I read about within the news. This is why I have to test it myself.This works great other that you can't enable the Beta with the camera blocked.
For example this article claims covering the camera didn't block the engagement of AP or FSD Beta.
Tesla Driver Monitoring Fails to Keep Driver Attention on Road - Consumer Reports
Tesla’s camera-based driver monitoring fails to keep driver attention on the road, Consumer Reports' tests show. Autopilot works with the driver-monitoring camera covered.
I'm not going to do the cell phone test as I feel like its low hanging fruit neural network wise. It's pretty easy to train a neural network to detect a cell phone or what looks like a cell phone within view of the camera.
What I'm concerned by is the lack of any notification or alert when using the UI.
When Tesla released the V11 UI "upgrade" they caused hundreds of thousands of drivers to struggle to find controls where driving down the road while on AP or FSD Beta. I imagine quite a few didn't have the self restraint to stop themselves from menu diving while hurling down the road at 70mph.
The very fact that I was able to figure out this wretched UI is both an indication that the driver monitoring sucks, and indication that I need to be a bit more self restrained when being annoyed that I can't find something.
So I made a rule like I've done thousands of times while driving to do a better job.
roblab
Active Member
Yeah, I drive with my hand (singular) at the 6 o'clock position, but I have to deliberately move the wheel now and then. I find it easier to forget AP and just drive it myself.
Fundamentally they all come down to accommodating different types of users.
I don't think either the V11 UI nor the Driver Monitoring were really tested all that well.
The V11 UI has been rather thoroughly ridiculed on user forums. It's not just change, but its change to something where there was no obvious benefit from it.
The Driver monitoring system has no way to audit it so we have no idea how well its actually working. I plan on trying to duplicate CR's setup by having a slider like they did. The reason for this is I want to see what it does if I cover it AFTER engaging AP or FSD Beta.
Once I know how how it responds to being covered then I can figure out if there is anything about me to explain why it doesn't seem to work.
Is there something about putting a bag over the cabin camera that creates a different result from a slider? Need to understand the details here and why results are inconsistent
For me its really about comparing expected functionality, and what actually happens.
If I put my thumb over the camera I expect FSD Beta not to engage.
If I put my thumb over the camera while driving down the road on FSD beta I expect it to complain rather quickly and to have to take over.
If I put my thumb over the camera before engaging AP on the highway I expect it not to engage
If I put my thumb over the camera while driving down the road on AP I expect it to complain rather quickly and to have to take over.
Why do I have to test these?
Functionality changes between firmware builds, and the HW on a vehicle.
I'm sure there is a logical explanation behind why CR got the results they did.
Owners themselves are dissuaded to test the driver monitoring out of fear of a strike out so we don't have an ample amount of evidence as to why there are differences. With the blindspot monitoring cameras we know what the root cause is of different visual results. We know it took Tesla quite a few revisions to get the HW right. so it wasn't impacted by the turn signal light.
I would rather not test covering the camera as it's not the actual test I want to do. It's only a ground truth test like verifying a device under test is plugged in.
Verify that the base functionality is working, and then verify that it can see my face.
sleepydoc
Well-Known Member
I’ve gotten nagged to ‘pay attention to the road,’ even when I’m looking straight ahead and doing just that. It also doesn’t seem to matter where I’m looking, I still get nagged to move the steering wheel every 10 seconds no matter what so while it may use the camera it’s not clear for what it’s using it; it certainly doesn’t seem to be for tracking driving attention.
And I’ll agree with @S4WRXTTCS - v11 is a wonderful example of how not to design a UI. Many times I’ve been thankful I have FSD so I can figure out the screen!
And I’ll agree with @S4WRXTTCS - v11 is a wonderful example of how not to design a UI. Many times I’ve been thankful I have FSD so I can figure out the screen!
PACEMD
Active Member
Wake up and smell the nags. Maybe a correspondence course in advanced human UI interpretation........
I received an FSD strike recently because the car insisted that I wasn't paying attention to the road, despite applying force to the wheel and looking dead ahead. I believe the camera was tricked by having the visor only partway down pointing toward me, as opposed to pointing toward the windshield. This despite it not blocking line of sight to my eyes. Since then, I take care with the visor and haven't had any other issues.
Similar threads
