Video analytics at the edge, an ideal technology for 5G cloud monetization

Creating a programmable software infrastructure for telecommunication operations promises to reduce both the capital expenditure (CAPEX) and the operational expenses (OPEX) of the 5G telecommunications operators. What is exciting to many of us who work in this space is that the convergence of telecommunications, the cloud, and edge infrastructures will open up opportunities for new innovations and revenue for both the telecommunications industry and the cloud ecosystem.

In this blog, we focus on video, the dominant traffic type on the internet since the introduction of 4G networks. With 5G, not only will the volume of video traffic increase, but there will also be many new solutions for industries, from retail to manufacturing to healthcare and forest monitoring, infusing deep learning and AI for video analytics scenarios. The symbiotic evolution of video analytics and edge computing provides opportunities for operators to offer new services which they can monetize with their customers.

To learn more about our Azure for Operators strategy, refer to our e-book, "A cloud for network operators."

Video analytics, edge computing, and 5G  

Our first public disclosure of real-time video analytics was when we characterized it as the killer app for edge computing. Collaborating with the City of Bellevue, Washington we pursued their Vision Zero initiative to conduct a pilot study for live traffic congestion and safety at the cities’ camera-equipped major traffic intersections. We hosted a traffic dashboard powered by our video analytics solution to detect automobiles, pedestrians, and bicyclists at Bellevue’s Traffic Management Center (July 2017 to November 2018). The dashboard also helped Bellevue transportation planners understand traffic patterns over extended periods of time and led to the creation and assessment of a bicycle lane on one of its main streets. The project was a big success and the city won multiple national awards for its vision and pilots on incorporating video analytics for traffic management.

In parallel, as 5G began emerging, telecom operators began making large investments in their network infrastructure, with the lion’s share of the network capacity planned for video traffic. Interestingly, the aspiration of operators to use their infrastructure for digitization of various industries is beautifully aligned with Microsoft’s own investments in edge computing and video analytics. Edge computing is the catalyst that is leading to the convergence of the two infrastructures, and video analytics is the perfect service 5G operators can host on their edge computing servers.

Illustrates that the investments of 5G operators and Microsoft are aligned

The figure illustrates that the investments of 5G operators and Microsoft are aligned.

A revenue opportunity for 5G operators and Microsoft partners with new-age applications

There are several good examples we can envision for 5G operators to monetize video analytics services. Consider traffic monitoring and accident-avoidance solutions in smart cities, similar to what we implemented in our Vision Zero work with the City of Bellevue. A related application is integrating with self-driving cars with real-time video analysis from their cameras. Furthermore, consider modern smart enterprises where end-to-end experiences with video analytics and mixed reality are a natural component of private 5G network solutions. Additional examples include managing machines and robots in connected factories, or customer demands and services in retail stores and restaurants, or pedestrian traffic in sports arenas. In all these cases, 5G operators, in partnership with System Integrators (SIs), can use Azure edge computing products and Azure Video Analyzer to provide innovative solutions. 

The figure illustrates the coming together of 5G Operators, Azure edge video services and Systems Integrators (SIs) to offer future video analytics services to various industries.

The figure illustrates the coming together of 5G Operators, Azure edge video services, and Systems Integrators (SIs) to offer future video analytics services to various industries.

Microsoft already has an ergonomic, untethered holographic device featuring enterprise-ready applications that increase user productivity across industries from manufacturing to education, the Microsoft HoloLens. Looking at the not too distant future, offloading video processing from HoloLens to a nearby Azure Edge over a low-latency, high-bandwidth 5G network represents yet another example of how operators can offer new products. Microsoft cloud gaming platform, xCloud, also comes to mind as it delivers next-generation global game streaming. Leveraging the power of low-latency, high bandwidth 5G networks, alongside live video analytics on edge devices, operators can support a significantly enhanced gaming experience.

How Microsoft’s advanced technology makes all this possible

Microsoft has invested many years into developing large-scale live video analytics systems. We have published research papers with substantial platform advances, have developed related products, and open-source technologies. For instance, Microsoft Rocket is an open-source platform, which enables the easy construction of video pipelines for efficient processing of video streams. Its cascaded video pipeline, when combined with Azure Video Analyzer, makes it easy and affordable for developers to build video analytics applications into IoT solutions. The combination of Azure Video Analyzer and Microsoft Rocket along with Azure Arc enables easy configuration of resource-accuracy tradeoffs and orchestration over a distributed edge-cloud hierarchy. Azure Video Analyzer and Microsoft Rocket achieve an order-of-magnitude improvement in throughput per edge core for video analytics without compromising accuracy, lowering the total cost of ownership (TCO) at the edge.

Privacy preservation has been a central pillar of Microsoft Rocket’s goal to democratize video analytics. We embraced edge computing as a natural ally to preserve privacy with techniques to transform video at the edge that prevented leakage of personal information in the analytics. We also rely on secure hardware to protect against snooping and provide confidentiality during the analytics.

Specific to 5G, we have also incorporated extensive network monitoring and adaptations for fault tolerance and load balancing in the video processing pipeline to handle dynamic network conditions that are inevitable in all wireless networks. Our system, which we refer to as edge video services (EVS), works well with heterogeneous edge hierarchies supporting diverse hardware. For this, we created new technology for computation partitioning and an inter-edge orchestrator. EVS partitions the computation to make the best use of the available hardware at the edge and cloud infrastructure, while also co-existing with other workloads on the edge, as captured in the figure below.

Illustrates how the edge video services (EVS) partitions the computation to make best use of the available hardware at the edge and cloud infrastructure, while also co-existing with other workloads on the edge.

The figure illustrates how the edge video services (EVS) partitions the computation to make best use of the available hardware at the edge and cloud infrastructure, while also co-existing with other workloads on the edge.

Tailoring Azure Video Analyzer for real-world operation over 5G networks

We have been evolving our systems through pilots with operators and 5G network equipment vendors. Our engagement with Telstra, a prominent Australian telecom operator, is an example of an operator who wants to light up EVS. As part of Telstra’s mission to build a connected future for everyone, Telstra adopted Azure Video Analyzer and Microsoft Rocket along with Azure Stack Edge and Azure Percept preview. By intelligently distributing AI across edges, the amount of data processed was reduced by 50 times, thus leading to better utilization of Telstra’s 5G network. Telstra is developing scalable, cost-efficient solutions that help its customers optimize traffic flow and increase construction safety.

In our collaboration with Fujitsu, we trialed a private 5G solution for monitoring parking lots by analyzing video feeds from Fujitsu’s smart wireless cameras. In order to build autonomous networks with minimal complexity, Fujitsu adapted Microsoft Rocket into their 5G infrastructure where Microsoft Rocket and Fujitsu’s RAN containers execute alongside each other on an Azure Stack Edge. Microsoft Rocket substantially lowered the compute and network demands while providing low-latency and accurate visualization of the parking lot’s occupancy.

In another example, in collaboration with academic colleagues at Princeton University, Microsoft developed the world’s first 5G-based multi-hop camera network. This relay-based camera network uses edge servers and cameras fitted with WiGig radios to create a fully connected millimeter-wave (mmWave) network. This then allows for efficient streaming and analysis of live video in areas where direct line of sight to the base stations is often problematic, as shown in this demo video.

Looking to the future

In the years to come, people around the world will access and use 5G networks every day. 5G networks will continue to provide value across industries, providing high-capacity and low-latency connectivity to support an abundance of complex and useful applications. At Microsoft, we believe privacy-preserving live video analytics applications are an ideal fit for 5G networks. Our research and innovation, outlined in this post, continue to move us forward by clearing a path for inventing the next generation of live video analytics applications that will revolutionize our world—making it safer, more efficient, and more entertaining. To learn more about our Azure for Operators strategy, refer to the Azure for Operators e-book.

Source: Azure Blog Feed

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.