When it comes to counting people—just counting people, not tracking Wi-Fi or Bluetooth signals from mobile phones to serve up advertising and to use as a proxy for crowd size—there are not many highly accurate, automated technologies. Andrew Farah wants to change that. The CEO of San Francisco startup Density believes his company has developed technology that can replace the handheld tally counter as the state-of-the-art technology.
Last year, Density engaged with a number of diverse organizations—Workfrom, a Portland-based startup that helps freelance workers find coffee shops offering good Wi-Fi and available seating; several gyms and workplaces on University of California campuses; and a homeless shelter—to run pilot tests and evaluate the first version of its IR sensor.
That sensor, which transmitted a pair of parallel infrared beams, was mounted, at roughly waist height, next to exit and entrance doors of each establishment. Each time a person walked through the door and disrupted the IR waveform, the sensor recorded his or her presence. The walker’s direction—and, therefore, whether the person was entering or exiting the building—was deduced from whichever of the two IR beams was disrupted first.
The technology worked, but had a number of limitations, such as the inability to isolate individuals when they walked through the doorway in a tight group. So Density recalibrated the technology and, on Wednesday, released its new people-counter sensor. The device still employs infrared technology, but rather than using parallel beams, it transmits a single 8-foot-high by 4-foot-wide cone of infrared light from its position above the door. The IR beam bounces off the ground, or off individuals moving through the doorway, and the sensor measures the amount of time that passes before the light returns to the sensor. Not only that, it derives a profile from the returning light, forming what Farah calls a topological map that is based on the distance from the sensor of each person in the doorway. The processor in the module uses a computer-vision tool called bounding boxes to singulate each human figure in the doorway. It does this using “tens of thousands of points of data in every frame, taking 20 frames per second,” Farah explains.
The animated gif below shows the processor tracking two individuals walking through a doorway at the same time, stopping to greet each other, and then departing. One person is entering, while the other is exiting.
Farah says the technology is very effective at differentiating individuals who walk through a door at the same time, even if they are of similar height. “Just about the only way we would miscount people,” he states, “is if they were twins conjoined at the head.”
The Density module continually updates the number of individuals who are inside a building or a room within a building, based on its ongoing analysis of the data from the IR sensor mounted above a doorway. (Inanimate objects, such as suitcases or shopping carts, are filtered out based on the lack of a discernable human head.) Each time the tally changes, the module sends the updated figure to Density’s cloud-based server, via the building’s Wi-Fi network or through an Ethernet connection.
From there, Density makes the data available to its customers—usually the facility’s owner or management company, or the business or organization that leases the space—through an application programming interface (API). A large facility with multiple doors can be monitored by daisy-chaining the tallies from multiple sensors, which are fed into a central receiver that sends the updated total to the server.
Density’s customers might use the data internally to understand foot-traffic patterns, or to analyze the most and least busy hours at a place of business, in order to improve staffing. For example, the ridesharing service Uber is using Density to track foot traffic at its New York City support center, where thousands of Uber drivers seek services each week. “They are going to use Density to determine busy times and quiet times, and predict spikes in visitation,” Farah says. :In order to improve a driver’s experience, they plan to more effectively staff their spaces and better understand how people use them.”
But some of Density’s customers are using traffic data as a service to their customers. Workfrom is integrating the number of likely available seats (based on the Density sensor) into its app, so that workers on the go can find the least-crowded spots.
Density is not creating its own application for consumers, because its philosophy is that consumers should be able to access the information through the services and apps they already use. “So if you use a mapping app [with business listings], or an app that rates restaurants,” Farah explains, you should be able to see how crowded places are through those apps. (He did not reveal the names of any particular mapping or recommendation services that Density is targeting.)
But reliable data on foot-traffic data could also be valuable information to some surprising stakeholders. “Some businesses have expensive insurance policies that set limits on how many people can occupy spaces that people visit regularly,” Farah says. If insurers could verify the number of people within a space over time, he adds, perhaps they might offer premium discounts for policy-holders who abide by occupancy limits.
According to Farah, a homeless shelter has also evaluated Density’s technology. The shelter is interested in using the traffic data to attract more grant funding, since some grant-making organizations like to see proof of how many people a shelter serves before awarding it funds.
Public-transit systems and hospitals are among the other types of organizations that Density thinks could benefit from accurate people-counting technology. A transit authority could use Density to understand when trains are very full, or when they’re practically empty. An ambulance service could use the data, for example, to determine which nearby hospital is the least busy before deciding where to bring a patient.
Anonymity and not relying on cellphones are two of Density’s main selling points, especially for businesses whose customers have responded negatively to technologies that leverage cellphones to monitor traffic. For example, Nordstrom took heat from its customers in 2013 after it began testing technology from Euclid Analytics, which collects MAC addresses transmitted by cell phones’ Wi-Fi signals to estimate traffic size and patterns. The retailer subsequently dropped the technology.
Density does not charge for the sensor module, which is powered via Power-over-Ethernet or through a wall outlet, though it does charge for access to the data. Annual plans cost $45 per month, while a month-to-month plan costs $95.
The company has also just announced that it has raised $4 million in Series A funding. The round was led by Upfront Ventures’ managing partner, Mark Suster, who is also joining Density’s board of advisors.