Inferencing at the Edge with HyperFlex

While at this year’s Cisco Live! Barcelona, our staff interviewed many of the amazing thought leaders who joined us to demonstrate their work with Cisco. We’ll be sharing those interviews, and short blog post summaries, for the next few weeks.

The first interview is about how Vispera, which makes visual intelligence solutions for retail, works with Cisco on inferencing at the edge with HyperFlex. As Erdem Yoruk, chief scientist at Vispera, explained, the company uses image recognition technology for its shelf monitoring and inventory management solution, Shelfsight. Cisco HyperFlex, together with Intel’s Xeon Scalable Processors, provides the compute power needed to perform inferencing at multiple retail locations.

As they stood in front shelves with cereal and other products, Yoruk explained how Shelfsight, the latest product from Vispera, uses a camera solution for real-time shelf monitoring and in-store management. Running on Cisco HyperFlex Edge servers powered by scalable Intel Xeon processors, Shelfsight takes measurements from the shelves every 10 minutes. It recognizes every product visible on the shelf down to a granular level, generating KPI metrics and notifying store personnel of any non-compliance issues so that they can take immediate action and replenish the shelves as necessary.

Vispera has spent the last two years focused on in-store fixed camera solutions, which is what brought them to the collaboration with Cisco and Intel by using on-prem high-performance computing. This unique collaboration is ultimately what led to the creation of Shelfsight. It’s an ultimate combination of tools and companies for on-prem AI computing that’s already live in multiple retails stores throughout Turkey and India.

Powered by Intel

As Sidney Cerqueira, an Account Manager at Intel, continued, the collaborative solution uses the same hardware as other enterprise applications, eliminating the need for costly specialized accelerators to run complex AI workloads. The second generation of Intel’s Xeon Scalable Processor has a new technology called Intel Deep Learning Boost, which accelerates the inferencing, and runs new instruction sets, called avx-512. This improves image recognition, object detection, language processing, and speech recognition. Running on two node Cisco HyperFlex Edge with the new generation of Intel Xeon Scalable Processors takes advantage of this new deep learning technology and allows new workloads like image recognition to be performed at the edge where the data is.

Tune in next week when we talk with customer David De Roock from gas shipping company Exmar about the challenges Exmar was facing, and why choosing Cisco HyperFlex helped solve his issue.