Hacker News new | ask | show | jobs
by ljlolel 776 days ago
In order to make vision-only depth sensing work well, you need ground-truth data. Buying a few thousand lidar to get a ton of ground-truth training data for the NN to learn distance on its own from vision is an obvious move. They're not switching to Lidar.

source: i debugged lidar and built localization and behavioral planning algorithms for self driving cars

4 comments

Yes, 10 years after the initial announcement, Tesla has just now got around to ordering some LIDARs for capturing ground-truth training data. They then went out of their way to get the one LIDAR made to look palatable to end customers instead of the superior industrial ones everyone puts 8 of on their recording cars.

Not very credible.

> Tesla has just now got around to

What's the source that they haven't ordered from Luminar previously? They have definitely used Lidar test rigs for years.

See sibling comment, this wouldn't be a first time occurance.
And they've been doing this for several years already, with Luminar hardware mounted on engineering vehicles with manufacturer plates.

https://www.cnet.com/roadshow/news/tesla-model-y-luminar-lid...

Why would they not simply use lidar in the cars themselves?

Besides, I imagine many of the issues with vision (e.g. adverse weather) could not be addressed with training-time data.

> Besides, I imagine many of the issues with vision (e.g. adverse weather)

LIDAR is bad in rain, fog and snow.

So are cameras.
And people
Because $1000 of Lidar hardware per car would be $1.85 billion per year.
If you make millions of them per year, lidar won't cost $1k per car. OTOH existing lidar designs are not designed for mass market manufacturing or serviceability.
Take out of his 56B comp package
Those are stock grants which don’t directly affect the company’s net income.
They do affect net income when they vest. It is treated as an expense in GAAP.
If you had multiple cameras for stereoscopic vision. Couldn't you figure out the depth data vision only without lidar? I'm sure there was a good reason Tesla didn't go with a stereoscopic camera system (at least front facing) from the get-go.. they already have 3 cameras up there with 3 FOVs. Could add another camera there in a stereo setup for depth data.
Isn't human stereoscopic vision only good for something like 20 feet or so? The cameras behind the rearview mirror are even closer together than that.