Intelligence at the Edge using converged edge systems allows critical insights to be extracted from video data right at the network edge, helping to speed reaction times, reduce the risk of data transfer, and drive better decision-making.

As the amount of video data generated daily from video sensors continues to climb, businesses are seeking faster and more effective ways to analyze it in order to extract real-time insights and drive better decision-making. A joint solution from HPE and NVIDIA® is bringing accelerated analytical capabilities of artificial intelligence (AI) and deep learning out to the edge, delivering video analytics to speed reaction times, reduce the need for data transfer, and mitigate the risk of data loss or corruption.

A growing number of video sensors are capturing more content than ever, with some industry estimates asserting that up to thirty billion images will be captured every second by the year 2020. Driven by the increasing need to monitor urban areas for abnormal activity and major leaps forward in recording technology, these sensors installed all over the globe are producing a mass of data that continually escalates in both size and complexity. The convergence of these trends is quickly resulting in a big data problem, one that leaves the industry seeking breakthrough new analytics techniques in order to properly collect, process, and analyze it.

AI and deep learning are helping businesses quickly and efficiently analyze these troves of rich video content, and these techniques have seen even greater adoption in recent years as new technologies come to market that provide a cost-effective and commercial-ready way of implementing AI/deep learning in the real world. For example, in 2012 the Google Brain Project used thousands of computers to form a large neural network that could recognize pictures of cats, but system performance and techniques have now advanced to the point where a device that fits in your hand can perform similar tasks. Today’s more advanced deep neural networks use complex algorithms, large amounts of data, and the computational power of NVIDIA graphics processing units (GPUs) to shift the paradigm, and as a result today’s computers are able to learn and react with unprecedented speed, accuracy and scale.

Speed reaction time for safety and security scenarios

Monitoring, reacting to and preventing safety and security incidents is one of the most prevalent use cases for video sensors. From businesses and individuals wanting to protect their personal property, to cities striving to provide a safe and smart environment for their citizens, analyzing video data is becoming a critical tool for identifying high-risk behavior, increasing understanding of past incidents, and ensuring public safety.

In safety, security, or failure prevention scenarios, the speed of reaction is one of the most critical factors to effectively addressing the situation. For example, the response to a severe accident at rush hour may involve immediately dispatching first responders, changing road signs to reroute traffic, and automatically archiving the accident footage as police evidence. In this instance, the accuracy and speed of the insight delivered by video analytics can be a key first step in driving the proper response, so video analytics applications must be capable of delivering immediate insight.

Intelligent Video Analytics (IVA) are also highly useful in scenarios where the objective is to filter out normal activity and bring just abnormal activity to the attention of humans. For example, while an algorithm can immediately determine that an accident has occurred, humans are still needed for subjective decisions such as whether the severity of accident warrants one ambulance or five. IVA can quickly provide insight that optimizes the time humans spend making decisions by bringing timely, prioritized information to light so people can make consistently correct and informed decisions.

Extract real-time information at the intelligent edge

Traditionally, the preferred way to process video content from sensors was to transfer it from the edge of the network (where the sensors and the entities they monitor are located) back to the data center for further analysis. But in scenarios where real-time action is paramount, the latency created by transferring data across the network means that by the time critical data reaches the core it is already outdated. Because of this, the optimal way to process this data is to do it as close to its point of creation as possible – the edge. Additionally, edge analytics also resolves concerns about economics and feasibility of large data transfer, security, compliance, data corruption in transit and duplication.

Innovations such as the new Converged Edge Systems product space created by HPE, are allowing critical insights to be extracted from video data right at the intelligent edge, enabling businesses to leverage the power of AI and deep learning to process rich content like video and extract timely insights from it without needing to transfer it to the core or cloud. Imagine the value of having a small piece of your data center that is fully functional at the edge, running the same unmodified analytics engine with new, real-time data it can act upon instead of outdated information.

HPE Edgeline Converged Edge Systems are a pioneering line of compact and ruggedized systems capable of high performance and precision data acquisition, running powerful enterprise class analytics and taking appropriate control actions directly at the edge, in real-time, while leveraging the full benefits of NVIDIA® Tesla®-accelerated artificial neural networks to enable capabilities like AI and deep learning. The HPE Edgeline is fully customizable, with system type, number of industry-standard server modules, number of NVIDIA GPUs, memory, and storage all able to be tailored to fit the precise number and quality of video feeds and intensity of analytics needed. NVIDIA® Tesla® P4 GPUs are purpose-built to boost compute efficiency for servers running AI-based analytics, capable of slashing inference latency by 15x in any hyperscale infrastructure and providing 60x better energy efficiency than CPUs. Download the NVIDIA IVA Infographic today to learn more about IVA applications.

Together, HPE and NVIDIA are offering edge system innovations that allow insight to be extracted right at the edge of the network. For more information about how our leading-edge video analytics solution can help your business achieve rapid time-to-insight, please connect with us on Twitter at @HPE_HPC and @NVIDIADC.