Published Apr 15 2020 11:15 AM 4,046 Views

NVIDIA's GTC DIGITAL 2020 event is a collection of online content that provides training, research, insights, and direct access to the brilliant minds of NVIDIA’s GPU Technology Conference.  Microsoft delivered a variety of sessions with a couple focused very specifically on IoT Edge applications using small form-factor GPU accelerated devices from NVIDIA's Jetson line of embedded devices.  These sessions include a presentation on the latest techniques for "Productionizing GPU-accelerated IoT Workloads at the Edge" and an interactive workshop demonstrating how to perform anomaly detection on multiple video input streams with IoT Edge, IoT Central, and Cognitive Services using a $100 NVIDIA Jetson Nano device.  The workshop content is highly recommended if you are looking for a hands-on introduction into these topics, in approximately one hour you can build a custom anomaly detection pipeline with edge-to-cloud reporting. 

The sessions are linked in the images below and available online for viewing at your leisure.  If you have questions, comments, or need help following the workshop material feel free to reach out to @pjdecarlo on Twitter for in-person assistance!


Productionizing GPU Accelerated IoT Workloads at the Edge.png


In this session, we will cover a variety of techniques, tools, and services to assist in the production of Computer Vision Solutions deployed in Edge environments that run on NVIDIA Jetson hardware.


Topics will cover:


  • Usage of Azure DevOps tooling and workflows to produce deployable container-based GPU-Accelerated modules for use with Azure IoT Edge
  • The Azure IoT Central - Software as a Service offering for fleet management and runtime-configuration of IoT Edge devices
  • DeepStream compatible custom object detection models built with the Azure Custom Vision AI service offering
  • Jetson-containers tooling for producing CUDA compatible container images and flashable CUDA-capable base OS images targeted to NVIDIA Jetson hardware.

The presentation will consist of live demonstrations of these topics and include resources for reproduction.  At the end of the session, you will leave understanding how to develop production grade GPU-accelerated container workloads backed by deployment, management, and configuration capabilities in Microsoft Azure.


Anomaly Detection Workshop.png


In this workshop, you'll discover how to:

  • Build a solution that can process up to 8 real-time video streams with an AI model on a $100 NVIDIA Jetson Nano device
  • Remotely monitor and operate your device using services and features in Azure IoT Central
  • Develop and deploy custom AI models to an IoT Edge device

With this solution, you can transform pixels from a camera into insights to know when there is an available parking spot, a missing product on a retail store shelf, an anomaly on a solar panel, a worker approaching a hazardous zone., etc.

We'll build this solution using NVIDIA DeepStream on a NVIDIA Jetson Nano device connected to Azure via Azure IoT Edge. DeepStream is a highly-optimized video processing pipeline, capable of running deep neural networks. It is a must-have tool whenever you have complex video analytics requirements like real-time object detection or when employing cascading AI models.

IoT Edge gives you the possibility to run this pipeline next to your cameras, where the video data is being generated, thus lowering your bandwidth costs and enabling scenarios with poor internet connectivity or privacy concerns. We'll operate this solution with an aesthetic UI provided by IoT Central and customize the objects detected in video streams using Custom Vision, a service that automatically generates computer vision AI models from pictures.

Version history
Last update:
‎Jul 09 2020 04:48 PM
Updated by: