What the MS camera can do

[lolachampcar]

This is a duplicate post but I thought it worth its own thread. Mods, please merge or delete if it has already been posted.

I spent about an hour with my dog on my lap (sound asleep) watching this. It was worth every moment although the second half wearable stuff was a bit dry.

Brains, Minds Machines Seminar Series: Computer Vision that is changing our lives - YouTube

Thanks to the TM poster for finding it.

 

[pgiralt]

That was a very cool video. Very exciting to hear about their next-gen technology with 8 cameras!

 

[Stoneymonster]

Very interesting. It seems like the current hardware is pretty underutilized and has capabilities yet to be exploited.

 

[jkn]

19:05 moose is starting to go across the road. In this situation car should stop immediately. Trying to go front of it could be deadly mistake. Typically moose does not turn back. It will try to go over faster. Because of long legs, in collision 500 kg of bones and flesh will hit windscreen. Often front of the roof is not strong enough. So people on front seats will die. I have seen seen, but not yet collided with moose.

Based on what I have read, MX is not designed for moose collisions.

 

[lolachampcar]

I believe what you saw was data collection while decisions were still in the hands of the driver.

This video dovetails nicely with information I got a few years ago indicating ME was already driving around US streets (including cities complete with stop and go and navigation though intersections) autonomously gathering data.

 

[Matias]

Very interesting. It seems like the current hardware is pretty underutilized and has capabilities yet to be exploited.

According to that presentation Tesla uses EyeQ3. It is unclear to me whether those very cool demonstrations (free space detection, lane detection without lane markings, pedestrian detection, debris detection etc.) were running on it.

I agree, that it seems that EyeQ3 is very powerful and can do a lot of things which are not yet implemented in Tesla. Very exiting!

 

[Stoneymonster]

According to that presentation Tesla uses EyeQ3. It is unclear to me whether those very cool demonstrations (free space detection, lane detection without lane markings, pedestrian detection, debris detection etc.) were running on it.

I agree, that it seems that EyeQ3 is very powerful and can do a lot of things which are not yet implemented in Tesla. Very exiting!

True, I thought somewhere in there he said it was EyeQ3 or strongly implied it was current hardware. But it could have been Q4 stuff.

 

[Sound's Good]

Have been working since '09 with gear designed to accommodate firmware expansion and truly, it's amazing to communicate with the manufacturer and individuals involved with the machines' creation and see users' feedback incorporated in sometimes minutes. This is what I envision for this machine too. Such fun!