After clicking “Watch Now” you will be prompted to login or join.
At-the-Edge Deep Learning & Traditional Machine Vision Integration: Ultrasonic Welding Inspection
Itai Shchorry, Optimal+ | Tobias Finzel, Vitesco Technologies GmbH | Michael Schuldenfrei, Optimal+
GTC 2020
Car manufacturers are rapidly adopting new and complex technologies in the race for electrified, connected and autonomous vehicles. The manufacturers of parts for these vehicles, led by large tier-1 suppliers, are pressed to provide bleeding-edge components at ever-lower prices, while ensuring that they never fail in the field. They are working hard to ensure their manufacturing lines are running as efficiently as possible, while producing perfect products all the time. In this talk, Vitesco Technologies (formerly Continental Powertrain) and Optimal+ will present their joint journey to implement an in-line solution using deep-learning on nVidia GPUs in a 24x7 production environment, enabling maximum efficiency and quality for an ultra-sonic welding operation. The solution combines traditional image processing with deep-learning inferencing at the edge to cover the full range of possible defects. We'll also show how the solution lays the foundation for additional applications, by providing a scalable, reliable platform to manage the entire lifecycle of AI solutions at the edge.