Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017” • Hacker News Board

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#51

Post by maxerickson » Sat, Nov 12, 2016, 7:36 PM UTC

LeifCarrotson wrote: post #32

maxerickson wrote: post #24

Earlier quoted context omitted.

Is 10 watts the correct figure? The powertrain uses 300+ Watts, so it seems like a manageable number if it is correct.

That's a big "+" on your figure - it's about 300 Watt-hours per mile , or 3600W constantly, if you're going 60 mph. For comparison, one horsepower is about 750 watts - a Tesla is a very low-drag car. Anandtech reported [1] that the TDP of the whole board is around 250W. The Tegra SoCs are probably around 10W, but that doesn't get you much in the way of GPU horsepower. 1: http://www.anandtech.com/show/9903/nvidia-anno…

Yeah, I screwed up the math.

Point stands though.

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#52

Post by bcantrill » Sat, Nov 12, 2016, 7:38 PM UTC

ilaksh wrote: post #35

bcantrill wrote: post #23

Could someone who understands this space weigh in on how technically interesting this is? (Or isn't?) In particular, their research paper on "End to End Learning for Self-Driving Cars"[1] seems to yield a system that requires an unacceptable amount of manual intervention: in their test drive, they achieve autonomous driving only 98% of the time. But I have no real expertise in this space; perhaps this result is impre…

It's incredibly freaking amazing if they are using deep learning to drive via mainly cameras only 98 percent of the time. No one else can do that. 98 percent is obviously a lot.

Thanks -- that answers the question! So fair to say that it's impressive because of the absence of LIDAR and/or other sensors -- and that by adding LIDAR to such a system one could presumably get towards 0% manual intervention?

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#53

Post by dogma1138 » Sat, Nov 12, 2016, 7:39 PM UTC

varelse wrote: post #21

NVIDIA's Drive PX 2 has too high power consumption and too low perf/W for the moment. They're winning this space because they are there more than that they are the best possible solution. And they may continue to win because successful execution of an 80% product is worth far more than a 90+% powerpoint processor cough TenSilica et al. cough , or because this is such a huge potential market, it might actually go to a…

>I'm not a big fan of $150K high-end servers filled with $5000 GPUs that can be bested with clever code on a $25K server fill with $1200 consumer GPUs. But I am a huge fan of charging what you can while you are unopposed. It's just that I think that state is temporary.

There is virtually not a single "enterprise" grade product which can't be made at least 50% cheaper (or sometimes 10 times...) with off the shelf consumer grade hacked hardware....

Enterprise products always have a pretty steep markup, but what you lose with those 1200$ GPUs is both features (e.g. virtualization, thin provisioning, DMA/Cuda Direct etc.) and support. When you buy a 5000$ CPU over a 500$ with the same performance what you pay for is reliability and support, if you don't care about that then fine, but when you need to launch a 100M$ service on top of that platform you won't really care about the price tag it's all in the cost of doing business.

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#54

Post by TaylorAlexander » Sat, Nov 12, 2016, 7:41 PM UTC

kchoudhu wrote: post #25

bluthru wrote: post #22

Earlier quoted context omitted.

What's a hard case for vision? That's how our eyes work. LiDAR has problems with precipitation too.

Depth map extraction from vision in real time depends on accurate algorithmic merging of past frames, color gradients and motion vector extraction to come up with a 3D map of what's around the vehicle. Contrast this with LIDAR, which can present a depth map in real time by sending out an array of light pulses and timing how long they take to come back to the IP. Which method is more likely to have implementation erro…

Humans don't use accurate depth maps to drive.

Machine learning is great at complex algorithms if done well.

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#55

Post by ChuckMcM » Sat, Nov 12, 2016, 7:47 PM UTC

tedunangst wrote: post #50

ChuckMcM wrote: post #45

Earlier quoted context omitted.

The argument reasoning I've heard goes like this; People drive reasonably well using vision primarily and with imperfect visibility of their environment. Computer learning networks can classify imagery at least as accurately as humans and sometimes more so. A computer using imagery that is well classified from an array of visual sensors with near perfect visibility should be able to drive as well, or better, than a h…

ML seems pretty bad at classifying things it hasn't seen before though. There are quite a few examples where an input outside the training data resulted in misclassification. Humans may not always see a white truck in a snowstorm, but is computer vision going to see it either? Or will it pattern match the few visible parts as something else entirely? Or dismiss the truck entirely as noise?

I don't disagree, both humans and ML are bad at classifying things they haven't seen before[1]. However that reasoning doesn't disqualify either vision only auto driving systems or machine learning.

Both statements are true:

"Computer driven cars may crash, even fatally, when they encounter a situation that they do not recognize." and

"People driving cars may crash, even fatally, when they encounter a situation that they do not recognize."

The success criteria for self driving cars is that they can drive at least as well, in the common case, as the set of human drivers who are defined to be "good" drivers. And self driving is not invalidated by a computer's mishandling of an event that a good driver would also mishandle.

I expect that self driving systems will be differentiated by how well they handle the unusual cases so a Mercedes system might do better in an unusual situation than a Chevy system. And all of this discussion is orthogonal to LIDAR :-).

[1] http://puzzlephotos.blogspot.com/

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#56

Post by ebalit » Sat, Nov 12, 2016, 7:54 PM UTC

tedunangst wrote: post #50

ChuckMcM wrote: post #45

Earlier quoted context omitted.

The argument reasoning I've heard goes like this; People drive reasonably well using vision primarily and with imperfect visibility of their environment. Computer learning networks can classify imagery at least as accurately as humans and sometimes more so. A computer using imagery that is well classified from an array of visual sensors with near perfect visibility should be able to drive as well, or better, than a h…

ML seems pretty bad at classifying things it hasn't seen before though. There are quite a few examples where an input outside the training data resulted in misclassification. Humans may not always see a white truck in a snowstorm, but is computer vision going to see it either? Or will it pattern match the few visible parts as something else entirely? Or dismiss the truck entirely as noise?

You can use a model that gives you its classification uncertainty. Bayesian SegNet for example [1]. We may also adapt the legislation for how vehicles should look like we did to make human driving easier (ex: tail and side lights).

1: https://arxiv.org/abs/1511.02680

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#57

Post by kchoudhu » Sat, Nov 12, 2016, 7:54 PM UTC

Animats wrote: post #49

kchoudhu wrote: post #44

Earlier quoted context omitted.

Aww yeah -- another Challenger! What team were you on?

I ran Team Overbot.[1] We were way overdesigned for off-road and underdesigned for going fast. [1] http://www.overbot.com

Same for us -- speed was not a consideration at all. We figured that if we could keep going and finish, we'd be one of the top 3 teams.

As it turned out, our mechanical engineering was great. What got us was software: we failed to free memory for passed obstacles, ran into memory exhaustion issues as a result and crashed out at mile 9. When we fixed the leak and reran the course, the truck finished in just over 7 hours.

It was an object lesson that garbage collection won't save you from memory-related issues. In retrospect, it was the also first time I ever encountered a million dollar bug. D'oh.

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#58

Post by mattnewton » Sat, Nov 12, 2016, 7:56 PM UTC

ChuckMcM wrote: post #45

Animats wrote: post #20

I'm still not happy with self-driving on vision alone, or vision augmented with radar. There are too many hard cases for vision. Everybody who has good self-driving right now - Google, Otto, Volvo, GM - uses LIDAR. Self-driving is coming to the first end users in 2017, in Volvo's test of 100 vehicles. Volvo has multiple LIDARs, multiple radars, multiple cameras, redundant computers, and redundant actuators. They're b…

The argument reasoning I've heard goes like this; People drive reasonably well using vision primarily and with imperfect visibility of their environment. Computer learning networks can classify imagery at least as accurately as humans and sometimes more so. A computer using imagery that is well classified from an array of visual sensors with near perfect visibility should be able to drive as well, or better, than a h…

Why not shoot for better than human performance and "cheat" any way possible along the way? To paraphrase a quote I can't remember by who, do we care if a submarine "swims"?

Besides, with even with lidar, the problem is hard enough.

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#59

Post by mattnewton » Sat, Nov 12, 2016, 8:00 PM UTC

dogma1138 wrote: post #53

varelse wrote: post #21

NVIDIA's Drive PX 2 has too high power consumption and too low perf/W for the moment. They're winning this space because they are there more than that they are the best possible solution. And they may continue to win because successful execution of an 80% product is worth far more than a 90+% powerpoint processor cough TenSilica et al. cough , or because this is such a huge potential market, it might actually go to a…

>I'm not a big fan of $150K high-end servers filled with $5000 GPUs that can be bested with clever code on a $25K server fill with $1200 consumer GPUs. But I am a huge fan of charging what you can while you are unopposed. It's just that I think that state is temporary. There is virtually not a single "enterprise" grade product which can't be made at least 50% cheaper (or sometimes 10 times...) with off the shelf cons…

I don't understand your argument, so maybe this is off base, but if you are saying people in industry aren't replacing their supercomputers with commodity gpu's, you're wrong; both apple and google have massive purchase orders for commodity nvidia gpus because they aren't just cheaper, they are better at this application. And I imagine other companies are as well.

Edit: "replace" is probably not the right word, this is work that the old systems don't do well, but they aren't throwing out x86 racks for gpus of course. It's just instead of buying more of the same for machine learning applications.

Re: Nvidia on new self-driving system: “basically 5 years ahead and coming in 2017”

#60

Post by kchoudhu » Sat, Nov 12, 2016, 8:01 PM UTC

TaylorAlexander wrote: post #54

kchoudhu wrote: post #25

Earlier quoted context omitted.

Depth map extraction from vision in real time depends on accurate algorithmic merging of past frames, color gradients and motion vector extraction to come up with a 3D map of what's around the vehicle. Contrast this with LIDAR, which can present a depth map in real time by sending out an array of light pulses and timing how long they take to come back to the IP. Which method is more likely to have implementation erro…

Humans don't use accurate depth maps to drive. Machine learning is great at complex algorithms if done well.

Do the newer systems assume network connectivity to a backend processing facility?

Personally, I view each additional layer (network, someone else's data center, machine learning) to be something that can fail and put people in danger. Assurance via local brute force is much more reassuring for me.