Autonomous Driving Meetup #7

Autonomous Driving Munich
Autonomous Driving Munich
Öffentliche Gruppe

Microsoft Deutschland GmbH

Walter-Gropius-Straße 5, 80807 München · München

Wie du uns findest

There are 450 parking spaces available on an outdoor parking lot (about 150 meters away) - corner "Lyonel-Feininger-Straße" / corner "Anni-Albers-Straße"

Bild des Veranstaltungsortes


It's time for another autonomous.driving.munich meetup after summer break! Join us for some intense discussions, great talks, snacks and drinks. This time we are hosted by Microsoft.

___ Agenda ___

_ Introduction & Welcome

_ Session 1: From „brainwave“ to FPGA – Microsoft take in the simulation of autonomous systems
Mitra Sinha, Kurt Niebuhr / both Microsoft (Azure Compute engineering)

Mitra Sinha and Kurt Niebuhr from the Azure Big Compute engineering team in Redmond will deliver the Microsoft keynote at the event – focusing on how Azure enables the training, inferencing and lifecycle management of leading edge AI algorithms like Deep Neural Networks (DNN) at scale. They will cover the whole spectrum of the Microsoft portfolio from the cloud, with services such as Azure Batch AI and the latest Nvidia GPUs down to Microsoft’s AI silicon innovation (e.g. “Project Brainwave”). Finally, the attendees will learn about Microsoft’s investment in software for the simulation of autonomous systems like the recent “” acquisition.

_ Dinner Break

_ Session 2: Building Artificial Brains for Autonomous Driving
Karim Mansour, Co-Founder and Vice President / SIGRA Technologies

Abstract: TBD

_ Session 3: What would you do with a hundred AVs?
A brief overview of building a testing environment for self-driving cars that scales
Rasmus Rothe, Founder / Merantix
Clemens Viernickel, Applied AI Strategy & Operations Lead / Merantix

Testing of machine learning based systems such as autonomous vehicles faces two key difficulties: Firstly, model progress is non-linear rendering traditional approaches such as using component level metrics ineffective to drive actionable development. Secondly, operating AVs is expensive and slow, which is why you need to make sure to get the most out of all the data you are recording. Scenario based testing describes an offline testing environment which evaluates the performance of AV stacks end-to-end based on a catalogue of thousands of short driving situations. This approach enables a semantic and scalable comparison between software versions.

_ Socializing

___ Further Details ___

_ The talks are all held in english and will take ~30 minutes each
_ Sponsor: Microsoft