Tempe Police Release Video of Uber Accident • Hacker News Board

Re: Tempe Police Release Video of Uber Accident

#461

Post by extrapickles » Thu, Mar 22, 2018, 1:46 AM UTC

eanzenberg wrote: post #181

pdonis wrote: post #168

Earlier quoted context omitted.

> I have better night vision than this camera, I feel as a human I would have seen the bike. I'm not so sure. Under conditions where there is low light everywhere, so the eye has time to adapt, human low light vision is very good. But under the conditions shown in the video, where there are street lights present and the pedestrian comes out from a gap between the street lights, your eye won't be dark adapted and you…

This is literally the reason for high-beams. If the road is not illuminated enough you turn those on and drive as fast as you are comfortable being able to fully stop if a deer jumps out of left field (or in this case, a pedestrian is walking across the road).

It's typically illegal[0] and otherwise a bad idea to operate high beams when there is other traffic around as you can blind other drivers.

[0]: Page 46 of https://www.azdot.gov/docs/default-source/mvd-forms-pubs/99-... Other states have similar laws, such as California: http://leginfo.legislature.ca.gov/faces/codes_displaySection...

Re: Tempe Police Release Video of Uber Accident

#462

Post by imh » Thu, Mar 22, 2018, 1:47 AM UTC

The video makes it look like the woman popped out of the shadows and makes it look like this was unavoidable. But that's not what a person would really see. People's eyes have great dynamic range. Take this picture for example (not mine):

https://cdn-images-1.medium.com/max/2000/1*OQtewILwLl-EssbWY...

What you'd see in real life is much closer to the edited version on the right, while unedited pictures (the version on the left or the uber video released) would make it seem like you can't see shit. A driver paying attention probably would have seen this person from far away. At the very least, the video doesn't convince me this was unavoidable.

Re: Tempe Police Release Video of Uber Accident

#463

Post by theptip » Thu, Mar 22, 2018, 1:48 AM UTC

blagie wrote: post #442

ttul wrote: post #315

Earlier quoted context omitted.

That would be criminal. Nobody at Uber would risk that.

I've seen criminal behavior at a much more reputable organization. Top-10 university, senior leadership. Much more criminal. This shit happens and it's more common than you'd think.

Sad but true.

Re: Tempe Police Release Video of Uber Accident

#464

Post by Maybestring » Thu, Mar 22, 2018, 1:48 AM UTC

ineedasername wrote: post #327

Okay, so Tempe police aren't going to be able to get & grok the raw data on how the AI was handling things. So how does that shake out in an investigation? It seems like they'll have to rely very heavily on the good faith of Uber (hah!) to honestly interpret & communicate the truth of that side of the story. Releasing the video so soon without the equally important (probably more so) AI side is almost deliberately mi…

The NTSB is opening an investigation, and Congress critters are watching.

The Tempe police aren't going to lead the crash analysis. If history is any guide, they'll probably have NASA doing code reviews.

Re: Tempe Police Release Video of Uber Accident

#465

Post by markc » Thu, Mar 22, 2018, 1:48 AM UTC

Watch the footage carefully and you'll see that the car actually tracks to the right immediately before impact.

When centered there's about 20" between the sides of the car and the lane markers. (120" lane - 80" car) / 2.

At the moment of impact, the victim appears to be between 3' and 4' from the right lane marker. To avoid the fatality, the car would have only had to track left 2' from center.

So my question is - what was the last possible moment such an avoidance maneuver (veer left) could have been initiated?

I think that's an interesting factor in assessing whether this tragedy was avoidable or not. Certainly seems like a technical failure here here in that the car was easing slightly to the right and not braking or veering left.

Re: Tempe Police Release Video of Uber Accident

#466

Post by jakelazaroff » Thu, Mar 22, 2018, 1:48 AM UTC

TylerE wrote: post #444

sitkack wrote: post #417

Earlier quoted context omitted.

1) It does NOT show what the LIDAR sees. 2) They should be using a dark adapted / night vision camera and probably are, the video feed presented is just that, the best one that "presents". 3) They should be using thermal infrared to spot living things 4) They were severely over driving their headlights if what we are seeing is "reality" as seen by the onboard computer. I want to know how much access to the hardware U…

It's Arizona. I just checked. The high today was 92F. Pavement would almost certainly be warmer than that. How's a thermal camera supposed to detect 98F on a background that is proably +/- 5 degrees of that?

That's the temperature at the peak of day, right? The temperature in the middle of the night, when the crash occurred, should be closer to the low.

Re: Tempe Police Release Video of Uber Accident

#467

Post by cmurf » Thu, Mar 22, 2018, 1:48 AM UTC

fudged71 wrote: post #251

I do wonder if perhaps Uber made the video darker before releasing. Or used a camera that is not the main sensing camera. Because I would hope the cars are smart enough to increase ISO to get more detail at night time.

I think it's possible to look at the video file and find markers for modification. Certainly this is the case for JPEG, TIFF, PSD, vs Raw files. It's not impossible but very impractical to render a Raw file, edit it, and then unrender it back into a Raw file. It's also pretty malicious and I think it would show a corrupt intent to modify a Raw image in this manner.

There are Raw equivalents for video, but consumer cameras are all post rendered, and heavily compressed. So modification is going to change all kinds of metadata, but in particular there will be more than the expected quantization from double compression.

Again, if that were demonstrated, I think it would show a corrupt intent. The intent to deceive.

Re: Tempe Police Release Video of Uber Accident

#468

Post by linkregister » Thu, Mar 22, 2018, 1:48 AM UTC

ashwinpp wrote: post #319

Apart from all the questions about the camera, resolution and culpability, we need to re-think what it means to be training and evaluation of autonomous cars with respect to the rare incidents. Training the cars based on data collected from roads is heavily biased towards incident free conditions. This does not give any training or feedback on the rare occasions such as these. If there was a learning algorithm decidi…

Luckily, self-driving companies agree.

All of the major ones use test tracks with live hazards, including the scenarios you mentioned.

Waymo hazard testing https://www.google.com/amp/sanfrancisco.cbslocal.com/2017/08...

Uber hazard response https://youtu.be/LM8Zw8fiPXE

Uber test track https://www.google.com/amp/s/amp.businessinsider.com/ubers-f...

Cruise Automation’s test track https://www.autoevolution.com/news/general-motors-starts-an-...

As far as driving on public roads, I tend to agree. As do the companies. They all have a safety driver behind the wheel, ready to take over.

There was clearly something that went wrong here in this tragedy, but it’s not a reckless disregard for safety on anyone’s part that led this to happen.

Re: Tempe Police Release Video of Uber Accident

#469

Post by djsumdog » Thu, Mar 22, 2018, 1:49 AM UTC

jeffmould wrote: post #135

jackpirate wrote: post #78

Earlier quoted context omitted.

It's clear from the video that a human driver actually would've had more trouble since the pedestrian showed up in the field of view right before the collision, yet that's in the visible spectrum. That's not at all clear to me. I don't know too much about cameras, but it looks to me like the camera is making the scene appear much darker than it actually is. In the video, you can see many street lights projecting down…

I have to agree. Just like a normal camera has issues in low-light, it is clear that this camera is diminishing exactly how light the road ahead was. While I can't say confidently that I would have been able to stop to prevent hitting them, watching the video in full screen does lead me to believe that I would have seen them and been able to apply the brakes at least enough to reduce the impact. Also, watching the vi…

But this has got to be just the black-box camera, right? Surely the actual camera they use as a driving sensor is much better than this? Not to mention the LIDAR and all the other sensors that should have caught this.

Re: Tempe Police Release Video of Uber Accident

#470

Post by sp332 » Thu, Mar 22, 2018, 1:49 AM UTC

sv12l wrote: post #265

atonse wrote: post #25

How did LIDAR and IR not catch that? That seems like a pretty serious problem. It's clear from the video that a human driver actually would've had more trouble since the pedestrian showed up in the field of view right before the collision, yet that's in the visible spectrum. When I argue for automated driving (as a casual observer), I tell people about exactly this sort of stuff (a computer can look in 20 places at t…

> It's clear from the video that a human driver actually would've had more trouble since the pedestrian showed up in the field of view right before the collision, yet that's in the visible spectrum. I'm not sure, pls look that pic https://imgur.com/a/VfBck , you can clearly see there exists at least 10-15 meters b/w them right at the time when she pops up. Now I don't know the speed of the car, but I'd wager, a human…

At 38 MPH, the car would cover that distance in 0.7 seconds. That is on the low end of human reaction times for braking, so an average person might not have time.