Our Blog

05

Jun

Vision Intelligence On the Edge

The combination of Microsoft and NVIDIA technologies enables Empired to build next-generation computer vision applications that run on the edge computing layer for our clients. Empired is currently demonstrating how video streams ingested by various cameras can be analysed at the edge in real-time.

Empired uses its proprietary approach to deep learning and leverages NVIDIA libraries and DeepStream running computer vision-based applications on several platforms including the NVidia Tesla line of GPUs and accelerators from the Jetson line.

Our use of Microsoft Azure and NVIDIA supports a diverse set of use cases that use AI to understand pixels and sensors and analyse metadata. It also offers the versatility to deploy from NVIDIA Jetson on edge to NVIDIA Tesla in Azure. Our 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.

Retail analytics and insights

The most recent upgrades to Azure and NVIDIA allow Empired to offer larger multi-GPU cluster or a microservice in containers. These offerings allow a highly flexible system architecture that opens new application capabilities.

Our Enablement programme caters for co-development with our clients where we make available our Docker images of pre-trained and supervised assets on DeepStream from NVIDIA GPU Cloud (NGC) and couple this with Azure Cognitive Services. When we combine this with Kubernetes, we see applications scale rapidly and enable our framework to support tens of thousands of video frames that can be ingested from hundreds of cameras.

We believe with Microsoft's new integration of NVIDIA and DeepStream with its managed edge offering, Azure IoT Edge, Microsoft is pulling away from the competition. In part, this is due to the strategy for integrating with user-centric ground-breaking solutions from such providers as NVIDIA; but also, the ability to recruit in these services elastically and not experience any split in deployment sets Azure apart. 

We can deploy Azure IoT Edge in a variety of devices including Jetson TX2 and the more recent addition, Jetson Nano. Because we build and deploy as a modular set of containers, this too allows for quick deployments and the ability to reach velocity faster by our clients.

Our team can acquire video frames from cameras and feed them to the containers for seamless utility, producing outputs which are inferenced by the neural networks running in our containers. These can also be fed as an input to additional modules such as Azure Functions or Azure Stream Analytics for further processing.

We use a wide range of fluid architectures to make Azure IoT Edge and DeepStream a potent combination to develop sophisticated computer vision-based applications running at the edge.

We see ourselves as an early mover in this space and have seen edge computing demands rise; it is becoming the destination for deep learning models running as a stand-alone. Video Intelligence is the most pervasive use case for the edge to date. When it comes to bringing this applied science to enterprise-grade solutions and providing enterprise level returns, we believe we are ahead of the curve.

Currently, we are working on building extensible capability and working with open source edge computing layers that will create more seamless and adaptable intelligent assets that leverage IoT PaaS running in Azure.

Posted by: Dr Steve Leven, Head of Advanced Technologies and Transformation | 05 June 2019

Tags: Machine Learning, Edge Compute, Intelligent Traffic, Retail Analytics


Related blog posts


Top Rated Posts

Blog archive

Stay up to date with all insights from the Empired blog