New Volta companion computer for Pixhawk brings computer vision, more

The DJI Manifold will run accelerated deep learning. But will the Volta ?

The DJI Manifold runs a tweaked version of Ubuntu and has a GPU supported by deep learning tools such as Caffe, Theano, Torch etc.
So it has a large amount of open source (The slide above puts a question mark). Also, the slide above does not show the DJI Manifold running MAVproxy. But as the Manifold runs Ubuntu, it will run MAVproxy.

A key question is whether the Volta GPU is going to be supported by the main deep learning tools. I can find no support from Caffe, Theano, or Torch for accleration via the MALI T628 GPU on the Arm chip used by the Volta. If someone else can see that, then I would love to know.

This DIY Drones discussion, is looking at how to put a GPU that supports accelerated deep learning in a plane or copter.

Here is a link to Nvidia's Drive PX2 presentation at CES2016 to show where the deep learning technology is moving in relations to driver assisted cars, which will move down to planes and copters as the tech is miniaturised.

Caffe | Deep Learning Framework

You need to be a member of diydrones to add comments!

Join diydrones

Comments

Nicholas Witham January 11, 2016 at 3:24am

@Aldo, but it has a cool looking case.
Aldo Vargas January 11, 2016 at 2:30am

Is this not just an odroid XU-4??

I have one and I payed 80 usd for it... http://www.hardkernel.com/main/products/prdt_info.php?g_code=G14345...
Serge Contal January 11, 2016 at 2:17am

Ouch, a fan ? Never ever ever use a fan in a small and dense product with LTE or other radios.

Come on, this is on board of a plane/multirotor, a simple heatsink should be enough given the airspeed that flows in the structures !
Developer

Pete Hollands January 11, 2016 at 1:00am

The DJI Manifold will run accelerated deep learning. But will the Volta ?

The DJI Manifold runs a tweaked version of Ubuntu and has a GPU supported by deep learning tools such as Caffe, Theano, Torch etc.
So it has a large amount of open source (The slide above puts a question mark). Also, the slide above does not show the DJI Manifold running MAVproxy. But as the Manifold runs Ubuntu, it will run MAVproxy.

A key question is whether the Volta GPU is going to be supported by the main deep learning tools. I can find no support from Caffe, Theano, or Torch for accleration via the MALI T628 GPU on the Arm chip used by the Volta. If someone else can see that, then I would love to know.

This DIY Drones discussion, is looking at how to put a GPU that supports accelerated deep learning in a plane or copter.

Here is a link to Nvidia's Drive PX2 presentation at CES2016 to show where the deep learning technology is moving in relations to driver assisted cars, which will move down to planes and copters as the tech is miniaturised.

Caffe | Deep Learning Framework

of 4

This reply was deleted.