Deploying Neural Networks in Embedded Applications

Hands-on with Nvidia Jetson

Deep learning is known for being power hungry and usually you need large graphics cards or a data center (cloud computing) during training. In embedded applications cloud computing or a large computer may not be available so you have to do the computation at the edge of the network (edge computing), and there are certain things to take into account when doing this.

Fredrik Molander from Nvidia will talk about deploying neural networks in embedded applications and you will get the chance to apply this using their embedded platform called Jetson:
In this lab you will deploy a deep learning application on a Jetson TX2. You’ll learn to customize and optimize neural network models for improved inference performance. You’ll work through hands-on exercises to:

• Acquire live images using the onboard camera
• Apply neural network models that detect objects, classify images, and perform image segmentation
• Customize and integrate your models into an embedded application using the TensorRT runtime

We have 15 Jetsons available so be sure to be quick to register. In case of many registrations we may organize so you have to sit two persons for each station.