This content was created by Robert Scoble. This content was uploaded by Robert Scoble. This is a YouTube video titled "A look at the 'eyes' of future robots with Stereo Labs" created in December 2015. It features a conversation between Robert Scoble and Cecile, the CEO of Stereo Labs, discussing their Z camera, a 3D depth camera designed to replicate human vision for applications like autonomous navigation and augmented reality. They cover the technology, market positioning, competitors, and future miniaturization of the camera.
So who are you?
Hi, I'm Cecile, CEO of sterelabs. And we make the Z camera.
Yeah, so this is a 3D depth camera, right?
Exactly.
And this is going to be really important in the next few years as these things start to shrink down in size and get put on glasses, like magically.
Right, Exactly.
So what is your camera for right now? Who do you think will buy, will need this camera?
Right, so we're targeting two applications, autonomous navigation and augmented reality. And as you said, DEPS is a very critical information for these two applications. And so what we've done here is actually reproducing how human vision works. So because we all have two eyes and. But if you look at cameras on the market, they all have one sensor, one lens. So we decided to bring human vision to machines. So imagine what you can do with. Oops.
Yeah.
So if you put this on a drone, a car or robot, it will actually see and understand, just like we do. So imagine the world you can build with this. So a drone, you can actually put a robot or a drone at home and it will understand how the space is organized. So it can see if there's a person, it can actually recognize objects. So yeah, you can do a lot of applications.
How much does this cost right now?
$449. So it's the cheapest depth sensor on the market that works indoors, outdoors, and sees up to 20 meters.
So sees up to 20 meters is important. So it's doing. What is it, spitting out light and waiting for it to come back. How does it actually see the depth?
So it has two eyes, just like you and me. So there's the left camera and the right camera. And there is a slight difference between these two cameras. And they see the same scene but with a slightly different point of view. And this is actually how our brain works. And you can actually understand the depths from these two cameras. So we don't project anything in the space. It just understands where the things are.
Which is why it only can see 20 meters. Just like our human eyes can't tell 3D ness further away.
Exactly.
Other than overlap, we can't actually see depth either. It's just you're turning this into a computer model as it goes down, you can sort of see that model here. What are we seeing here? Explain what you're seeing here.
You actually see for every pixel of the image, you know where it is in terms of distance. So we can actually say, okay, this wall is at 5 meters away. And. And so you can actually measure the Whole space. Because for every pixel you have in your image, you know exactly where it is in space.
And why is this going to be important? Drones you sort of mentioned and self driving cars, it'll be important. But why is this really going to be important in the next decade?
Because look at the robot, you want him to drive autonomously, but today it's actually blind so it doesn't know where to go. If you give him this sense of depth, he can actually make decision and actually be smarter and say okay, I cannot go there because there is someone or I should go there because there's a door. So you are actually bringing eyes to machines so that we can interact with machines and they can actually help us, like following us, helping us. So yeah, we're actually bringing them the eyes.
That's it. Tell me about the competitors market for this. You know, Prime Sense got bought by Apple and was building a little 3D sensor, but it wasn't doing it really with two lenses. It was doing it by spitting out some infrared light and measuring how long it came to come back. And come back.
Exactly.
Tell me about how this fits into the market. Because there's Matterport, there's other companies out there.
Yeah, so it's. The main difference is really that we're. So the camera works on a very simple principle, human vision. So other sensor you mentioned, they worked on projection of IR or structure, light, time of flight, we all know them. There are good cameras for just to control maybe scanning objects. But as soon as you want to go outdoors and as soon as you want to have larger environment, you cannot use them because first they're blind by sunlight and then they have a very short range. So they're good for some application. But as soon as you want to do augmented reality outdoors in large space, you cannot use them. And as soon as you want to go outside, make an autonomous drone or robot, of course it's outside so you cannot use them. And in this case you would use a lidar which is extremely expensive. So.
And it's a moving part where this has no parts.
Yeah, yeah. For maintenance. It's very difficult.
Well, it's going to be interesting to see what you guys do. How did you fund this company? How did you get.
So we started five years ago and we actually developed a 3D camera for the movie industry. So we actually worked for five years in Hollywood and we developed a powerful software that replicates human vision. And we actually worked with the James Cameron team to build the smartest 3D camera for Avatar SQL. And then at the end of 2014, we said, hey, everybody needs to see in 3D, so we should bring everything we've done for the movie industry and bring to everyone in the world. So this is when we decided to launch the Z camera. So we launched in May and it's having, it's having a huge success towards developers and anybody who wants to build a new system, you know.
So, yeah, Google has this Tango phone coming out. Are you aware of that?
Yeah, of course.
How will this compete with that? Or is it a different system? Explain where you think the market will go over the next year.
So it's. Yeah, because I expect these.
I've seen like the intel camera has seven sensors on the back of the camera. I don't know how it's working. So you probably know that better than some potential competitor. Right?
Yeah. So the thing is, we really wanted to make it simple. So we gather everything we've done before into a very cheap and actually simple to use system. Because we really want to enable people to put this on any type of machines. So it's a bit different because we want it to be universal. Like it's just plug and play. You plug it and then your robot knows where to go. You plug it into your smart glasses and then it captures the space to the scanning and then add elements into it. So we strongly believe that it needs to be very simple to be universal.
Yeah. This is obviously going to get miniaturized over the next five years. How aggressively do you think it will get miniaturized? When is this going to be built into a glass that I'm wearing like a magic leaf, because it's fairly big today. You're not going to put that on your head.
Exactly, no.
Maybe a minor lid.
It really depends on the application. Today we're targeting autonomous navigation. So it needs to be bulky and strong to be put on a robot and it needs to be far to see far away. So next year we're going to come up with a new camera which is going to be half the size of this one. And then in two years we'll definitely see it even smaller. So it's very simple. It's just the distance between the two eyes that can be reduced.
Very cool. Where do we get it?
You get it online. Can go to sterelabs.com and you can order it and we ship right away. Wordworthy.
