At the GTC 2019 conference in San Jose, Microsoft and NVIDIA demonstrated how video streams ingested by multiple cameras could be analyzed at the edge in real-time. This partnership enables NVIDIA and Microsoft communities to build next-generation computer vision applications that run on the edge computing layer.
NVIDIA has built a set of libraries called DeepStream SDK for running computer vision-based applications on a variety of platforms including Tesla family of GPUs and accelerators from the Jetson family.
According to NVIDIA, DeepStream SDK supports a diverse set of use cases that use AI to perceive pixels and sensors and analyze metadata. It also offers the flexibility to deploy from NVIDIA Jetson on the edge to NVIDIA Tesla in the public cloud. The SDK lets developers integrate the edge to the cloud with standard message brokers like Kafka for large-scale, wide-area deployments. This combination is ideal for building applications such as retail analytics, intelligent traffic control, automated optical inspection, freight and goods tracking, web content filtering, target ad injection, and more.
The most recent version of DeepStream SDK can be deployed as part of a larger multi-GPU cluster or a microservice in containers. This allows a highly flexible system architecture that opens up new application capabilities. Developers can download the latest Docker images of DeepStream SDK from NVIDIA GPU Cloud (NGC). When combined with Kubernetes, the applications scale rapidly to support tens of thousands of video frames acquired from hundreds of cameras.
Microsoft is integrating DeepStream SDK with its managed edge offering, Azure IoT Edge. When it comes to edge computing platforms, Microsoft is definitely a prominent player. As an early mover in the segment, the company built an extensible, open source edge computing layer that is seamlessly integrated with its IoT PaaS running in Azure.
Azure IoT Edge can be deployed in a variety of devices including Jetson TX2 and more recent addition, Jetson Nano. Since the platform is built as a modular set of containers, DeepStream SDK can be quickly deployed as a custom container module. Developers will be able to acquire video frames from cameras and feed them to the container running the DeepStream SDK. The output which is inferenced by the neural network running in the container can be fed as an input to other modules such as Azure Functions or Azure Stream Analytics for further processing. This architecture makes Azure IoT Edge and DeepStream a potent combination to develop sophisticated computer vision-based applications running at the edge.
Edge computing is fast becoming the destination for deep learning models running in a stand-alone mode. With video analytics being the most powerful use case for the edge, Microsoft is partnering with NVIDIA to leverage their investments in DeepStream SDK.
