Startup Deploys AI-Enabled Vision to Deliver Clicks to Retail’s ‘Bricks’

Motionloft’s computer vision tools offer insights on retail and city planning.
by Scott Martin

Amazon knows exactly how many seconds you pace around their site while shopping for running shoes. But brick-and-mortar retailers stumble when it comes to measuring foot traffic.

Motionloft promises to put them in step.

The San Francisco startup offers a device that sports a specialized video “super sensor” and NVIDIA Jetson-powered AI for monitoring pedestrian, vehicle and bicycle traffic. It provides brick-and-mortar tracking — similar to online analytics — for these kinds of real-world traffic.

Customers of Motionloft, which was founded eight years ago, can analyze their retail foot traffic and vehicle numbers and understand how long those people or vehicles pause in a location. These analytics provide a window into unique offline metrics that are standard in online marketing data.

That’s because Motionloft is using NVIDIA’s edge-to-cloud Metropolis platform, making it easy to create smart city applications by transforming video into valuable insights.

“We focus on how vehicles and people move around in our world. We take the physical world and turn it into digitized data,” said Joyce Reitman, chief executive of Motionloft, a member of the NVIDIA Inception program.

Holy Grail for ‘Bricks’

Such smart cities data is something of a holy grail for the likes of marketers, retailers, real estate businesses and urban planners.

Restaurants can use it to see how quickly customers in line are being served, so they can measure service quality and staffing needs. A fast-food outfit might learn how to better staff peak hours to avoid long lines that would otherwise cause people in search of a quick bite to leave empty-handed. “Millennials want their food quickly,” said Reitman.

Real-estate investors can use the data to study foot traffic for property acquisitions. Urban planners can study needs for cities. Las Vegas is using Motionloft to measure the activity of vehicles, bicycles and pedestrians and provide the data to those interested in investment in economic development projects.

It promises a boon for online-to-offline marketers: Businesses can use such foot traffic metrics to gauge the effectiveness of digital marketing campaigns.

Customers of Motionloft include the cities of Boston and New Orleans. So is Tanger Outlets, a real estate investment trust with holdings that span 42 shopping outlets malls and more than 400 branded factory outlets across 26 states.

Motionloft devices are installed in more than 1,500 locations across the U.S., Canada, Japan and Guam.

Super Sensors and Jetson

Motionloft devices pack a ViMo “super sensor” that uses two cameras and the NVIDIA Jetson AI computer to take in and process image information. Using Motionloft image recognition algorithms, the software can identify and measure the activity of people and cars and provide “heat maps,” “dwell time” and “path tracking.”

“We do that with a sensor based on the Jetson technology. We design and manufacture the sensor,” said Reitman. “It’s a powerful combination.”

All of the personal data is anonymized and none of the raw footage is stored but instead is converted into metadata and color-coded visual representations of foot traffic flows and the like.

Motionloft trains its neural networks and prototypes enhanced code in the cloud on AWS with NVIDIA GPUs and constantly adds more data for training its algorithms — which results in a constantly improving product.

“Our customers have never before been able to get these kinds of insights,” said Reitman.