At the Conference on Neural Information Processing Systems (NIPS) earlier this month, Audi generated a lot of buzz with a presentation of its neural network, which generates highly precise 3D models. It’s trick tech, and it’s the brains behind Audi Traffic Jam Assist.
Audi became the first automobile brand manufacturer to showcase at NIPS back in 2016, and this year, the Audi Electronics Venture project team presented a 3D mapping system which uses a single camera. But Audi isn’t going about this ambitious project alone. It’s a collaborative effort using intel provided by tech companies within Silicon Valley, Europe, and Israel, with the the goal of developing fully autonomous cars.
Right now, the team at the Electronics Research Library at the Audi base in Belmont, CA is already using AI—with a reliance on data from the environment and cameras—to steer and park cars on highways and parking lots. Here, lateral guidance is provided 100 percent through neural networks. But you don’t have to be a researcher to see some of the team’s handiwork in action, because the Audi A8 is the first car in the world to make it to Level 3 autonomous driving (SAE), and the AI Traffic Jam Assist can currently drive in traffic up to 37 mph.
Obviously, the critical component is its refined AI, which allows it to be incredibly precise. First, a traditional camera captures a 120 degree radius in front, delivering 15 images a second at a 1.3 megapixel resolution. After they’re captured, the images are processed within the neural network, and this is where the magic happens. That magic actually being something called “semantic segmenting.” This is how each object within range is identified and distinguished as an approaching (or reversing) car, sign, signal, person or any other identifiable object.
Wish you had attended NIPS? So do we! But you can do the next best thing and test drive an Audi A8 and Traffic Jam Assist at your local AutoNation Audi dealership! Check out the video below for an overview!