Magic Leap Acquiring Dacuda's 3D Scanning Assets

[scolaner]

Magic Leap, the secretive MR startup with a sky-high valuation, is acquiring 3D scanning software company Dacuda. Or at least, parts of it.


Magic Leap Acquiring Dacuda's 3D Scanning Assets : Read more

 

[bit_user]

After the Google investment, I assumed it would just use Tango. Maybe they just want the patents, for strategic reasons, and Dacuda's SLAM team as a quick way to add skilled & knowledgeable workers to their staff.

 

[scolaner]

After the Google investment, I assumed it would just use Tango. Maybe they just want the patents, for strategic reasons, and Dacuda's SLAM team as a quick way to add skilled & knowledgeable workers to their staff.

I bet Google thought that too.

On the other hand, as I understand it, Tango and SLAM are quite different beasts. Tango has multiple components (some specialized) plus software. SLAM just uses any smartphone camera plus software. They also have slightly different applications, too.

 

[bit_user]

On the other hand, as I understand it, Tango and SLAM are quite different beasts. Tango has multiple components (some specialized) plus software. SLAM just uses any smartphone camera plus software.

Tango is a software stack that sits atop various hardware to provide SLAM features in a consistent API. It's not tied to any particular CPU, GPU, or sensor hardware, as they've demonstrated by getting it running on x86 & ARM CPUs and Nvidia, Intel, & Qualcomm GPUs. Sensor hardware they've used includes ToF, structured light, and stereo.

In point of fact, they're only now reaching the level of exposing full SLAM. Previously, they've not exposed much in the way of Mapping. The API mostly just exposed Localization, giving you the device's pose (location + orientation). They added 3D reconstruction, only somewhat recently. The latest release includes tantalizing floor plan mapping and 3D mesh extraction features. Here's the full release history:

https://developers.google.com/tango/release-notes

Tango is fine for smartphones and tablets, but should be great in an AR HMD, as people have been experimenting with. On the Google+ Tango feed, someone recently posted this AR visor to hold Lenovo's Phab 2 Pro.

https://plus.google.com/+AsierArranz/posts/Hz5EtWFJSWs

BTW, if you want to see some things people are actually doing with AR, the Google+ feed is not a bad source. Takes a while to sift through, however.

 

[bit_user]

Speaking of their Google+ feed, this post is worth checking out, for some dirt on Magic Leap:

https://plus.google.com/108763673144961418891/posts/XUvrDRKAcRt

I always suspected that realtime lightfield rendering had to be at the very edge of technical feasibility. Especially in a portable device.

The source said that developers do not have the necessary hardware to begin making Magic Leap software, and that means it will be hard to ship Magic Leap with any applications this year.

This is all the greater reason they should just use Tango. Tango apps are up on Google's Play store, today.

But their investors are probably worried that supporting Tango will turn their device into a hardware play. They know the big money is when you take first rate hardware and use it to leverage yourself into all kinds of new markets, like Apple has done with the iPhone. In order to do this, you need to own the application stack - not be "just another" Tango device.

Venture capitalists are in it for a 10x return on investment. They regard anything much less as a failure. However, $1.4B is a lot of money, and if they can even get a 1x return, that's got to be an option they can't disregard. But, if they thought they had the next iPhone on their hands, you can see why they'd pump so much cash into it.

 

[Craig_67]

That is a DODOcase SMARTvr in the photo

 

[bit_user]

That is a DODOcase SMARTvr in the photo

Yeah, I think Magic Leap hasn't officially released any photos, because here's the leaked prototype that's started the recent buzz:

http://www.businessinsider.com/magic-leap-photo-leak-prototype-2017-2

I don't really see the problem. Looks to me like a Nvidia Parker development board, integrated into a VR backpack. IMO, if their tech works reasonably well, I don't see it as a dealbreaker if they need a backpack for the first gen.

The thing is, Nvidia Tegra dev boards are always about that size, but then the chips get integrated into tablets and such. The point of a development board is to be adaptable to any application, so it has every conceivable connector and peripheral on it, plus debug headers and extra space to work.

Plus, there's the fact that they say it's not even a production prototype, but rather a data collection test rig. I guess we'll see a real production prototype (with a belt pack), soon enough.

 

[computerguy72]

I saw the backpack photo the other day as well. That Tegra box could end up smaller than the box for the Nvidia Shield. So it's my guess it's clocked high and needs that big heat sync as presumably it is running un-optimized code. They might also be anticipating a later chip but testing with an overclocked existing model. The primitive nature of it's looks is probably not an issue.

 

[bit_user]

I saw the backpack photo the other day as well. That Tegra box could end up smaller than the box for the Nvidia Shield. So it's my guess it's clocked high and needs that big heat sync as presumably it is running un-optimized code.

Look, it's mostly likely just a generic board straight out of the Tegra development kit:

http://www.nvidia.com/object/jetson-tx1-dev-kit.html

They might also be anticipating a later chip but testing with an overclocked existing model.

They said it's a data acquisition system, to gather video & sensor data for the benefit of their algorithm development. In this case, they probably just took what they had lying around and slapped together something they could use for this purpose.

For context, check out the businessinsider link, and read the article below the image.

What will be really interesting is to see the prototype with the belt pack(s). The article was published on the 10th, and they said the meeting would be "next week". So, it might've already happened. But they said it was a board meeting, which is non-public. Who knows if those will leak, but it's debatable whether they have anything to gain by releasing them.