Brilliant read. I knew they were using pixel maps and comparing with motion to arrive at at 3D world, but Lance's article was very in depth. Thanks for posting.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Autonomous Car Progress
- Thread starter stopcrazypp
- Start date
-
- Tags
- Autonomous Vehicles
diplomat33
Average guy who loves autonomous vehicles
Tech Crunch has a review of the Zoox robotaxi:
techcrunch.com
Zoox robotaxi with supercar DNA rolls toward SF, Vegas, Seattle streets | TechCrunch
The Zoox robotaxi has more in common with a supercar than you think. Look past the Zoox’s breadbox-shaped exterior and inside is the same carbon fiber
techcrunch.com
diplomat33
Average guy who loves autonomous vehicles
diplomat33
Average guy who loves autonomous vehicles
Mobileye is rolling out their SuperVision system.
They have released Highway Assist which is L2 hands-free on the highway to the zeekr001 in China last week. And looks like next OTA will be the full L2 hands-free "door to door":
Watch Highway Assist handle different highway scenarios:
Read more: Mobileye and Zeekr OTA Update Opens a New Chapter in Advanced Driver Assist | Mobileye Blog
They have released Highway Assist which is L2 hands-free on the highway to the zeekr001 in China last week. And looks like next OTA will be the full L2 hands-free "door to door":
Watch Highway Assist handle different highway scenarios:
Read more: Mobileye and Zeekr OTA Update Opens a New Chapter in Advanced Driver Assist | Mobileye Blog
Last edited:
I thought the shown BEV pixel maps were synonymous with the "bag of coins." Didn't Telsa moved away from the "bag of coins" in FSD v10+ to direct generation of a vector space by the NNs from the stitched 8 cameras?
Last edited:
AlanSubie4Life
Efficiency Obsessed Member
I'm a total novice in this area, so I can't speak with authority. Based on reading Lance's article you're right, that Tesla used to use a bag of points for creating the BEV, but has since started migrating to using a transformer in the NN. From Lance's article in reference to using bags of points for BEV:
Autonomous vehicles need to know the distance to obstacles in order to drive. Unlike LIDAR, cameras do not provide the distance (or “depth”) of each pixel. This is a well-studied challenge for camera-only perception. At CVPR 2020, Karpathy mentioned that Tesla used pseudo-lidar for this task: regress self-supervised depth for each image pixel, project it to a 3D psuedo-lidar point cloud, and then use any mature detector that operates on LIDAR point clouds.
At AI day, Karpathy suggested that this approach is not good enough: lane lines regressed in the images (top) look noisy when projected into birds-eye-view (BEV, at bottom) because of error in the depth estimate of each pixel.
To address this limitation, Karpathy presented a transformer that maps pixels between the camera input and BEV output, an interesting departure from convolutional architectures (CNNs). Transformers have clear appeal: they operate on sets whereas CNNs require a fixed grid (e.g., image) input and they are general architectures whereas CNNs have inductive bias baked in (e.g., we construct CNNs to generate multi-scale features across a gradually expanding receptive field). As Yannic Kilcher argues, this inductive bias may constrain performance in the large-data regime and Tesla has a lot of data.
"Bag of coins?" - that must have been some sort of Freudian slip.Bag of Dogecoins?
Yes, a well-understood reference in the high-tech world, captiring a high level of future performance uncertainty and fluctuations in price.Bag of Dogecoins?
diplomat33
Average guy who loves autonomous vehicles
diplomat33
Average guy who loves autonomous vehicles
This is raw data from the InnovizTwo lidar. Incredible resolution!
diplomat33
Average guy who loves autonomous vehicles
Interesting video on the state of self-driving cars. It focuses on Cruise but it presents information about self-driving in a very easy to understand way. The main point of the video is that the focus right now is on L4 robotaxis and L4 trucks, not L5.
Here are the chapter times:
00:00 Self driving cars are finally here
02:15 Levels of self driving, explained
03:32 Where are we now on self driving cars?
05:07 Thank you Masterworks!
06:10 The big fight for the future of driverless cars
07:04 Interviewing Kyle Vogt, CEO of Cruise
08:20 Robotaxi meets world (and police)
10:00 The two near futures of self driving
11:52 The long term future of driving
Here are the chapter times:
00:00 Self driving cars are finally here
02:15 Levels of self driving, explained
03:32 Where are we now on self driving cars?
05:07 Thank you Masterworks!
06:10 The big fight for the future of driverless cars
07:04 Interviewing Kyle Vogt, CEO of Cruise
08:20 Robotaxi meets world (and police)
10:00 The two near futures of self driving
11:52 The long term future of driving
I'm excited for the future of AVs!!! What I'm nervous of is humans - we fear change. Remember the uproar when Uber disrupted the taxi industry? Robotaxis are going to disrupt taxis and gig-workers. I'd imagine the taxi companies (Yellow Cab, Checker, etc.) would start buying robotaxis and just get rid of their drivers, but you'd still be paying the cab company for your ride. Otherwise, they're done. I'm sure we'll see an uproar in the trucking industry too. Teamsters and other driver unions will no doubt lobby heavily to keep their industry afloat.
diplomat33
Average guy who loves autonomous vehicles
Oh no doubt. It happens with every new tech. But eventually the new tech takes over.
That is also why AV companies need to do their due diligence with safety and need to be transparent with the public. The groups who are against AVs will use any accident as a reason why we should not have AVs. And lack of transparency will be used to spread mistrust of AVs. Now, we can certainly discuss how certain AV companies may not be doing very well on the safety or transparency parts but they should be important priorities nonetheless.
Innoviz Two Spec.
0.05°x0.05°
Maximum Angular Resolution (HxV)10, 15, or 20 FPS
Pre Configurable Frame Rate0.3m-300m
Detection Range120°x40°
Maximum Field of View (HxV)
diplomat33
Average guy who loves autonomous vehicles
Ford pushes a software update that adds its L2 hands-free highway ADAS called "Blue Cruise" to 50,000 more vehicles:
www.teslarati.com
Ford adds BlueCruise to 50,000 cars for hands-free driving
Ford has now shipped its BlueCruise hands-free driving unit to around 66,500 Mustang Mach-E and F-150 vehicles, it said.
www.teslarati.com
diplomat33
Average guy who loves autonomous vehicles
I found an unedited video of an entire driverless ride in the Waymo I-Pace in SF:
The Waymo driving looks really good in this video. The Waymo Driver reminds me of a cautious and competent human driver.
The Waymo driving looks really good in this video. The Waymo Driver reminds me of a cautious and competent human driver.
This video showed up in my suggestions a few days ago and I passed thinking it was probably junk. Today at lunch it was there again and I clicked on it to give it 2 minutes. Wow it was well done and informative.
Similar threads
- Article
