ISJ Exclusive Interview: Wayne Arvidson and Greg Moya, Dell Technologies

ISJ speaks exclusively with Dell Technologies’ Wayne Arvidson and Greg Moya about how the company is leading the industry with an outcomes-based process for computer vision.

Organisations from every industry are in a digital race to turn data into business outcomes, faster. Computer vision is playing a more central role in that effort by accelerating time to value with its ability to connect video and other data at the edge with broader centralised networks.

From situational awareness to tracking customer flow in retail to optimising supply chains, the opportunities to improve outcomes using computer vision are immeasurable. Progress, however, is often impeded by operational and design complexities. How does an organisation match the right systems to the right workflows, and subsequently, those workflows to achieve the desired outcomes? That’s where Dell Technologies comes into play.

The company has been at the forefront of the enabling technology of computer vision. Combining with key industry stakeholders and an extensive partner network, Dell Technologies has developed and refined an award-winning process that takes customers from ideation to full-scale implementation, faster and with less risk.

Dell Technologies’ Global Industries team includes computer vision subject matter experts in just about every major industry. This team works with customers and partners throughout the process to provide end users with a Validated Design Solution that they can use to deliver real time business insights.

International Security Journal spoke to Dell Technologies’ Wayne Arvidson, Director Market Development & Strategy and Greg Moya, Global CTO about how their process for computer vision delivers optimised business outcomes for their customers across five key areas: Personnel and facility security, people experience, operational efficiencies, sustainability and enhancing revenue.

Prioritising outcomes over platform

“We’re trying to get away from the platform discussion,” said Arvidson. “We focus on outcomes and bring in the voice of the customer. We follow a unique validated design solution process and then take that back out to the customer. Through this process we’ve reduced the risk. We’ve sped up the implementation and delivered the workflows that the customer needs.”

“These outcome categories will entail different workflows depending on the industry in question. We can do object detection, which will help us identify abandoned objects in an airport terminal, debris on the runway or link a particular item or bag to a specific person.

“We can also do more complex workflows like monitoring ramp operations, ensuring the right personnel and equipment is in place to minimise aircraft turnaround time. For retail, we can perform analytics that show us which items customers examine, how much time they spend at the display, if they ultimately purchase the product. We can even plot the most common paths through the store so we can correctly place the highest margin goods in high-traffic locations, thus maximising per customer revenue. Operational efficiency is key, because the workforce is shrinking, not growing.”

“The foundation of what we deliver is the ability to ingest visual data once and then provide multiple insights for any one of five outcomes.”

Building computer vision workflows

Dell Technologies has derived these five outcome categories based on thousands of customer conversations. “These outcomes hit the nail right on the head in terms of what the customer is trying to achieve,” explained Moya. “All discussions start off with the customer’s current physical security infrastructure. But, then they evolve into other potentially beneficial outcomes.”

“The foundation of what we deliver is the ability to ingest visual data once and then provide multiple insights for any one of the five outcomes,” continued Arvidson. “It’s the ingest point for visual data. Ingest once and use that data in as many ways as possible to deliver the desired output.”  

Having laid the foundation for safety and security, Dell Technologies works with organisations to define the specific workflows that will achieve the outcomes an organisation requires. Defining these workflows, in turn, helps identify the components of the solution needed to achieve all these goals.

“The workflow defines which partner solutions should run on the Dell Technologies platform,” said Arvidson. “Then that drives the Dell Validated Design Solution that gets delivered to the customer.” The workflow determines the partner ecosystem; Dell Technologies first ascertains the workflow that can achieve a desired outcome, then matches the partner or partners that will provide that. The next stage is taking the partner software into the Dell Technologies lab to be validated against the hardware infrastructure.

“The output is a Dell Validated Design Solution,” added Moya. “This is the Dell platform including multiple applications that have been tested together. We produce a design guide with sizing and configuration information to help a customer, integrator and OEM replicate exactly what was done in the lab. This process is all about eliminating the risk inherent with delivering a complex deployment. It is an area that really differentiates Dell Technologies from other vendors in the market.”

The importance of lab validation

“We start by talking with our customers and understanding what outcomes they are trying to deliver,” explained Arvidson. “Those outcomes entail certain workflows and those workflows require certain algorithms and functionality. We look for the best of breed software partners that have the capabilities to support these workflows and assess them from both a technical and business perspective.”

“Once we identify the right partners, we start the lab validation process,” said Moya. “This is the most comprehensive in the industry and supported by over half a million square feet of lab space, dedicated to computer vision applications. We start by leveraging our partnerships with technology partners like Intel® and NVIDIA. We maintain mutual labs where these applications are brought in and tested on the various Dell platforms.

“Not only are they tested for basic functionality, but our partners provide developer support to help them optimise their code on our platform. A great example is when one of our analytics partners came to us. Initially, they were able to support seven streams per GPU on a server and after the lab validation process, they could do 70. That dramatically reduced the cost per channel for users. It opened up the market for the partner. Optimisation is critically important.”

“Not only are they tested for basic functionality, but our partners provide developer support to help them optimise their code on our platform.”

“Then, once the application is optimised, we go on to test the video management and analytics apps together, in real world scenarios that range from tens to thousands of cameras,” continued Arvidson. “This is where we look at performance metrics that are critical to the development of our sizing tools, ensuring that when we do a design, it works from day one and beyond and that we know how to scale it.

“That is complemented by our ‘test to fail’ process to ensure that we have the bandwidth required no matter what the situation. This process then drives an implementation guide so that when Dell or one of its OEM or systems integration partners stands the system up, you get the same results we obtained in the lab.”

“From there we build out our live and virtual demos with our Customer Solution Centres and then provide support for proof-of-concept designs, either virtually or onsite,” noted Arvidson. “This allows a company to bring its data into a secure environment and get some time with the various applications, seeing how they’ll behave in their environment, on a variety of Dell hardware. All of this is backed by a team of industry experts who can provide insights about specific use cases that have been implemented across the globe, at facilities large and small.”

Advantages of a ‘universal platform’ approach

“What we’re providing is not just an application,” said Moya. “It’s a ‘universal platform’ approach to accommodating diverse applications to achieve multiple, specific, and desired outcomes.”

Arvidson added: “If you look at what typically happens in the industry, it’s fragmented. Vendors often just test an individual application from a single partner. This is what I call ‘data sheet validation’. They test one thing, they test another thing, they test another thing and then assume those three things should work together. We found in the real world that isn’t the case. When we test things together, we help ISV partners uncover issues in their application or their coding. That impacts interoperability.”

“At the end of the day, what customers are buying is a workflow, not an application,” described Arvidson. “That workflow may be delivered through multiple application partners because some are good at one algorithm, and some are better at another. Some are better at real time and some are better at a more retrospective type of analysis. But you must deliver a workflow so we’re testing all the applications that deliver that workflow together to understand how they interact and how they need to be sized and configured to deliver the outcome that the customer can use.”

This process and the resulting workflow helps drive down costs. Because customers have tested the solution with their own data, they know it works. The validation process ensures optimal speed with less compute, and the removal of uncertainty of a failed implementation.

“We’re creating validation, best practices plus sizing,” said Moya. “We’re helping customers accelerate the adoption of innovation. A lot of times people are like deer caught in the headlights with this ‘art of the possible’. They’re looking at all this innovation and asking how it applies to them. We show how it applies to them by testing it at scale, testing it to fail and building it from an ecosystem of partners that they’re partnering with in the field.”

“People were stunned by the art of the possible,” declared Arvidson. “They didn’t know where to start because they were looking at individual use cases. But after all these thousands of customer conversations, we were able to create these themes. That’s where it became real for customers because they realise they must protect the people in their facility and we show them the workflows to do that.”

“Traditionally, Dell has been viewed as a hardware company with technologies that manage data,” concluded Arvidson. “We’re not just hardware and we’re not just managing data. We’re providing the resources that help you translate what’s possible with data. Ford doesn’t manufacture data, Ford manufacturers vehicles. An airport doesn’t manufacture data, it gets people from point A to point B. And that’s what we’re focused on – delivering the outcomes that matter most to businesses through this platform, through process, through our Global Industries division. That is our key differentiation.”

This article was originally published in the special September show edition of International Security Journal. To read your FREE digital edition, click here

Discover Dell Technologies’ solutions for computer vision workloads here: https://infohub.delltechnologies.com/t/computer-vision/