Nvidia Wants To Make All Cars Autonomous

Whether you like it or not, autonomous cars are coming - and Nvidia just made it a lot easier for manufacturers to jump on the self-driving bandwagon. The company's Deep Learning Institute (DLI) is offering advanced hands-on courses to aid in the development of autonomous vehicles. Provided it has enough expertise and money, any company can now build one. Ulp.

For the past year, Nvidia's DLI has been running training events and courseware for academic institutions, companies and government agencies. Its latest instructor-led workshops include advanced hands-on training to develop autonomous vehicles using the DRIVE PX 2 platform and DriveWorks software development kit.

The available courses cover hands-on experience using TensorRT to optimise, validate, and deploy a trained neural network for inference in a self-driving car application, image classification and object detection with Nvidia Jetson TX2 or TX2 developer kits and an introduction to DriveWorks with an emphasis on integrating DriveWorks modules into your custom code or applications.

Obviously, the above workshops are chiefly aimed at car manufacturers who are already dabbling in AI and wish to further develop their concepts. The DRIVE PX 2 platform, which is roughly the size of a lunchbox, is capable of performing 24 trillion deep-learning operations per second. This allows it to process up to 2800 images per second using a neural network-based algorithm - resulting in safe, unassisted driving.

There's also an intensive five-day training course available for major auto companies where Nvidia's certified instructors literally teach participants how to build a self-driving car on site. Currently, the DriveWorks courses are not available in Australia. (Not that we have an auto industry to take advantage. Tch.)

Lifehacker attended GTC 2017 as a guest of Nvidia.


Comments

    As part of autonomous vehicles, I'd imagine that protocols are needed for automated vehicle to vehicle communications of speed, direction, braking, near future manoeuvres (e.g. I need to turn left soon), and for the road to talk to the car (e.g. traffic slowing 1km ahead).

    The learning video was nice, but seemed a little light on detail, identifying a very light truckette as a 'truck' with the image of a 15t lorry. The issue for software is, I expect, they would have different vehicle dynamics and stopping distances, if that was part of the AI.

Join the discussion!

Trending Stories Right Now