This post has been republished via RSS; it originally appeared at: New blog articles in Microsoft Tech Community.
Computer vision at the intelligent edge is real and it is here! And not just computer vision—but high-performance, low cost computer vision thanks to NVIDIA DeepStream and Azure IoT Edge. Access sample code to get started and watch a live demo on the Channel 9 IoT Show for inspiration. Read on for more details.
Thanks to a collaboration between NVIDIA and Microsoft, the NVIDIA Metropolis video analytics application framework, which runs on EGX, is now optimized to work with Microsoft Azure IoT Edge. The NVIDIA Metropolis framework includes the NVIDIA DeepStream software developer kit.
Turn your installed RSTP cameras into sensors
With Azure IoT Edge and NVIDIA DeepStream, you can take a small, inexpensive NVIDIA Jetson Nano™ Developer Kit and analyze HD video streams in real-time. You can add an Azure IoT Edge device in the field to turn your RSTP cameras into sensors for IoT applications in retail stores, warehouses, manufacturing facilities, connected buildings, urban infrastructure, and more.
Why process video on the edge?
With all the power of the Cloud, why process video data on the edge? There are two reasons. One is that raw video footage consumes a lot of bandwidth if you send it to the Cloud. Secondly, you may want to filter and process video at the edge due to privacy concerns. It becomes efficient and cost effective to put an intelligent camera at the edge when installing new equipment. Or you can install a new intelligent gateway running Azure IoT Edge and NVIDIA DeepStream when you have cameras already installed and process data there.
Lowering the cost of computer vision
The Jetson Nano is a great device to process a handful of video streams. At approximately $100USD, the Jetson Nano can run multiple neural networks in parallel for applications like image classification, object detection, segmentation, and speech processing. It’s ideal because of its price point, AI capabilities and, of course, it runs Azure IoT Edge.
With smarter algorithms, you can squeeze more video streams from your hardware. That’s where the DeepStream toolkit comes in. It helps you set up your analytics pipeline. You can easily configure DeepStream to efficiently ingest as many video streams as you need.
In the live demo, using DeepStream, you will see that the Jetson Nano really shine processing 8 30fps video streams concurrently in real-time.
DeepStream also allows you to layer multiple AI models. For example, if you are analyzing video data from a construction site, you can use one model to identify a work person and you can use a second model to confirm whether the work person is wearing a hard hat and safety gloves.
Azure IoT Edge and NVIDIA DeepStream complement each other
Azure IoT Edge and NVIDIA DeepStream are complementary. Azure IoT Edge brings connectivity, security and the ability to deploy containers and manage them at scale. NVIDIA DeepStream brings video analytics performance, modularity and the ability to output messaging. Together they allow you to bring low-cost, high-performance computer vision to the edge.
Get started: sample code and live demo
Check out the sample code on this GitHub repo.
Watch the demo on the IoT Show and learn how to run your own AI models across multiple cameras in the most efficient way possible.