Audi demos all-round orientation via AI

December 04, 2017 // By Christoph Hammerschmidt
At the Conference on Neural Information Processing Systems (NIPS), one of the world's most prestigious AI conferences, Audi will be demonstrating how AI can be used to create a high-precision 360° model of the vehicle environment. Audi technicians use a camera for this purpose. The interesting thing about it: It's not a stereo camera but a simple mono camera.

A prerequisite for automated driving is a precise representation of the vehicle environment in real time. A key technology for generating this environmental image is Artificial Intelligence. A project team from Audi subsidiary Audi Electronics Venture (AEV) is presenting a mono camera at the NIPS that uses AI to generate a high-precision 3D model of the environment. A commercially available front camera serves as sensor. It captures the area in front of the car at an angle of about 120 degrees and delivers 15 frames per second with 1.3 megapixel resolution. These images are then processed in a neural network. Semantic segmentation also takes place there. Each pixel is assigned one of 13 object classes. This enables the system to identify and distinguish other cars, trucks, houses, lane markings, people and traffic signs.