The restart is underway, and as stores reopen, retailers must ensure customers and employees maintain adequate social distancing. To help track and manage those efforts, startup Voxel51’s platform can monitor in-store traffic and identify when locations are busy. The company’s metric, referred to as the physical distancing index, uses video feeds and computer vision modeling to reflect the amount of human activity in a particular location over time.
Jason Corso, co-founder and CEO of the technology firm, says physical distance measurement wasn’t on the radar before COVID-19. “Our vision as a company is to be the video platform where people go to rapidly build applications that need image and video capabilities,” he says.
When the pandemic hit and people began staying home, the team at Voxel51 looked for opportunities to help those around them. Corso says the initial goal of building a physical distancing engine to sit atop the core underlying platform was to provide better public awareness. “It was a tool to give people information that they’re not alone.” An analysis of human traffic patterns in various city hubs provided early insight into the pandemic’s social impact.
Now shoppers are venturing out again. “The retailer wants to keep people in the store, but they want to do it safely,” Corso says. Voxel51 can help brands stay ahead of potential problem areas without adding door counters and other labor hours.
Corso says they envision a dashboard for retailers, with a macro view of people in the store as well as parking areas; individual aisles and departments could also be monitored for traffic and densities, with the system updating about once a minute in many cases. “If they need to make on-the-fly changes or remediation to what’s happening, the retailer could do that,” Corso says.
He anticipates the technology will provide valuable data to help consumers decide when and where to shop to minimize their exposure. “If I’m going to a store, I’d like an accurate tool I can look at on my phone that shows things like density of people at the store as a function of time,” he says, emphasizing the predictions are based on historical data. Shoppers can plan their trips when it’s less likely to be busy. “That would be a way for people to feel more comfortable going to the store,” Corso says.
Data privacy concerns are addressed through a variety of protocols. A multi-layer approach to privacy preservation and security means nothing related to identity recognition occurs during video feed processing, and the technology manipulates image pixels to obscure personally identifiable information. “Our platform is best-in-class in doing that, and we take a conservative approach to what we redact,” Corso says. “We do the whole upper portion of the body and not just the face, for example, to be sure.”
While video has long played a role in retail, Voxel51 brings new capabilities around traffic analysis. “The PDI is a novel way of summarizing an awful lot of information that I don’t believe is done in many retail offerings,” Corso says. The technology also boasts potential uses beyond social distancing. “We can train models to determine if someone is interested in an item based on their posture, their gait and how long they’re spending at it.”