Hacker News Reader

Comment by tim333 | Hacker News Reader

>obvious orange obstacle

Apparently they loose the colour info and process black and white because it is easier. https://electrek.co/2017/05/16/tesla-autopilot-2-0-can-see/

This seems to have been an issue in some of the crashes and maybe was a mistake. See also for example the trailer an earlier Tesla crashed into https://www.theregister.co.uk/2017/06/20/tesla_death_crash_a...

Kind of obvious in colour but at the time it was probably a grey shape against a similar brightness grey sky

Also the big red fire truck http://autoweek.com/article/technology/ntsb-probe-autopilot-...

View on HN

Note that it's not (only) about colors... Human vision has trouble picking out colors in the dark, yet it still allows us to process world around us to a surprising degree. However until the "brain" of the car is good enough, it is simply irresponsible of Tesla not to use all the data available, from as many different sensors as possible. Testa should use color. And they should use Lidar.

Yeah I guess the perception has to be good enough to not crash in situations where humans wouldn't. It seems not quite there with tesla. Waymo seems to be better in that regard.

on human eyes, I was under the impression that we had a much better dynamic range than the best cameras on the market.

My understanding is that the eye doesn't have a great dynamic range, but our brain tricks us (as with a lot of vision). When looking at dark things, we can't make out bright detail and vice versa. As soon as you look at a bright area to make out detail, the eye adjusts, just like a camera.

> Apparently they loose the colour info and process black and white because it is easier. https://electrek.co/2017/05/16/tesla-autopilot-2-0-can-see/

If that's true, it looks like a very big mistake to me, considering that we use distinguishable color a lot to signal risks.

I (and I'm sure many others) did this for my Ph.D. project (group robotics) because it was the only way to get anything close to an acceptable frame rate on real time video using the ARM boards available. Colour processing is a big overhead, and the higher the definition of the camera, the worse it gets - so there's a definite trade off between the extra information from colour, and getting more frames/s to process. When you're extracting information from movement - i.e. the differences between frames, colour isn't what you're looking at in any case.

But knowing all that, and now knowing Tesla made that short cut, there is no way on the face of this earth I am ever driving a Tesla, or for that matter, driving a car anywhere in the USA that has them on the road.

You guys have a really good train network, right?

There are Teslas in many countries around the world, and autopilot is marketed as a feature in those as well.

About the US train network, it's good for freight but passenger trains are slow.

I've had my suspicions but this confirms it. Tesla's autopilot cannot detect non-moving obstacles. This thing is a killing machine! The problem is not caused by an opaque machine learning network running wild or anything of similar complexity. It's something as basic as a lack of sensor input. With the current hardware on Tesla cars full autonomy is simply not possible. It's basically a glorified lane keeping assistant with marketing that fools people into risking their lives.

Tesla's stance to blame the driver without admitting that their hardware is inadequate will back them into a corner in the long term. They cannot transition from highway to city driving without significant changes to the autopilot software and hardware and this means they effectively have to start over from scratch. They have to start working on a serious competitor to Uber or Waymo now otherwise Tesla will be too late to the market.

If that's true, that would be cutting the corners to a criminal degree. "Safety Red? Signal Orange? Traffic Yellow? Nah, we're good with Meaningless Grey, Meaningless Grey and Meaningless Grey."

Reply on news.ycombinator.com