Deep Learning for Self Driving Cars With a 2.3 Teraflop Motherboard, NVidia’s Drive PX

Jen-Hsun Huang, CEO and Co-founder of Santa Clara based GPU manufacturer Nvidia presented at this week’s GPU Technology Conference in Silicon Valley, explaining how deep learning will allow manufacturers to develop self driving cars and talks specifically about their Drive PX platform which will allow for sophisticated structure-from-motion and advanced stitching for better image rendering.

Huang believes that the big bang of autonomously driving cars is about to happen. Whilst driving is a learnt behaviour where understanding the environment is key, augmenting advanced driver assistance systems (ADAS) with deep learning neural networks are the comparatively sane alternative to writing an infinite number of ‘if.. then.. else’ lines of code.

In this talk Huang shows off the DARPA (funded) Autonomous Vehicle (DAVE) which comprised a deep neural network to train itself in the behaviour of driving and navigate in a hostile environment with no programming involved and avoid obstacles on one CPU with 225,000 labelled images.

Nvidia’s Drive PX is a 2.3 teraflop supercomputer which is able to perform 3,000 times quicker than the CPU which was used in DAVE and has twelve cameras which feed into it in order to train the network and mimic human level of learning to drive at a reported $10,000 for the developer kit which will be available in May this year.




About Gary Donovan

Machine Learning and Data Science blogger, hacker, consultant living in Melbourne, Australia. Passionate about the people and communities that drive forward the evolution of technology.
Show Buttons
Share On Facebook
Share On Twitter
Share On Linkedin
Share On Pinterest
Share On Stumbleupon
Contact us
Hide Buttons