10 Questions to a founder : Placemeter

We built cities around transportation, rivers, cars, train but little did we think about the human aspect.  With 3.5 billion people living in urban areas today and expected to double by 2050, it is becoming imperative that we change. Startup Placemeter offers the ability to quantify and analyze what we do, how we do it, when and how frequently using any cameras ( security, cell phone, IP cameras)  placed all around our cities. Using computer vision, it can quickly quantify population circulation so that city planners of tomorrow, retailers, architects,  can better design a world at human dimension. We sat down with Alexander Winter, co-founder and CEO of Placemeter to dig deeper.

A little about you, what is your background?

I was born and raised in France but am now a proud Brooklyn resident.

Prior to Placemeter, I co-founded LTU Technologies, a venture-backed startup and pioneer in large-scale image search by content. Before LTU, I was a researcher in Computer Vision at INRIA. I also worked at Airbus, building vision based missile guidance systems.

I hold an MSc and a PhD in Image Recognition from Telecom ParisTech, holding 9 patents in the field. I also advise various startups in New York and am a mentor of the Techstars NYC program.

Placemeter was conceived out of the desire to help people and businesses better understand and use the world around us.

After moving to New York City, the first thing my kids wanted to do was go to Shake Shack. We waited in line for an hour, then 20 minutes to get our burgers, and while waiting I realized they had a camera called the Shake Cam (publicly available) so I went home and started to analyze how the line moves based on my background in Computer Vision. It turns out the line doesn’t move at the same speed all the time – it depends on the number of counters open, etc. – I found it was possible to predict the best time to go.

When I moved to New York, I also realized the city is so dense, and there is an information imbalance – some people have data about the way cities are used, but it’s been in the hands of few rather than many; Placemeter wants to make this data accessible to anyone who wants to improve their environment.

Explain Placemeter. What does it solve?

The world population is going to gain 3.5 billion people in the next 30 years – the average city population is going to double – and cities are not ready for that. We need to optimize the way cities work, but to do that, we need to measure and quantify the way people use our physical world.

Data-driven urban design is making a resurgence in America and there is a push to have data behind everything people do and make – and there’s no way to do that without technology.

Wifi pinging is one way to do this, but it’s not very accurate. You can use satellites, but resolution is low and you can’t get 24-hour coverage. Physical sensors (loops in the ground for example) do the job but are single task and heavy to deploy.

At Placemeter, we believe that computer vision and video is the only scalable way to do that, which is why we’ve built an urban intelligence platform to quantify the movement of modern cities, at scale.

Placemeter believes that the more access to data, the better the city will operate and that a collaborative effort between both citizens and cities is essential to realize the vision of smarter, more accountable cities. This data will fuel innovation to help businesses, residents, and governments better understand the world around them and conversely, help cities be more cognizant of these institutions.

How does it work?

Through computer vision, Placemeter measures and quantifies the activity in our city’s places, streets, and neighborhoods. We apply the data to index the physical world, in real-time.

To quantify the activity in the world, Placemeter implements video sensors in people’s homes, businesses, restaurants that are facing outside, streets or public places. These videos are automatically analyzed to count people, vehicles, bicycles and other things, track their speed and their destinations.

Placemeter can count people coming in and out of stores or buildings, and infer how many people are inside, by only looking at the outside.

Placemeter’s algorithm then acquires, processes, and analyzes these real-world images to produce this data, and subsequently deliver information and predict patterns in cities.

Placemeter can also predict activity—activity is correlated with weather, holidays, time of events, days of the week, time of the day, calendar information, public video streams, crowdsourced data, etc.

90% is Placemeter computer vision technology. What did you have to solve before you could launch into the market?

For those who are not familiar with it, Computer Vision is a technology that mimics the way the human visual system works by emulating how your eyes and brain process visual input, to give computers a sense of sight. An image is a big matrix of pixels that doesn’t make any sense to a computer at first. In order for the computer to understand what is in this image, you train computers to see just like we teach children. It’s a matter of hard coding certain concepts into the system, and then supervised training, or trial and error at very large scales.

What makes Placemeter’s computer vision algorithms special is that for the first time we’ve made it automated and scalable. We are point-and-shoot computer vision. Everything is automatically calibrated; this is what makes it truly scalable.

In addition to solving for automation, we trained our system to work regardless of lighting changes, weather, sensor quality, frame rates, resolution, and viewpoint.

Finally, we designed our algorithms to process trillions of data points simultaneously. Computer vision is most often done on photos, but we’re the first to do it with videos at scale.

What is your answer to those who are concerned about privacy?

At Placemeter, we work every day to turn visual streams into actionable data for improving public spaces. Whether in business districts, traffic corridors or neighborhood parks, new video technologies help us understand, manage and improve these vital places.

We design our systems to yield useful data without compromising an individual’s privacy.

Placemeter’s innovative computer vision technology counts the number of pedestrians, cyclists, cars and other vehicles on a street or sidewalk in real-time video, and aggregates that data without using identity-detection technology.

As we design and continue to improve our systems, we adhere to the following principles:

Data without identity. We use computer vision algorithms to analyze video for the volume, speed and trajectory of pedestrian & vehicular traffic in cities. We don’t identify specific individuals with our computer vision technologies.

Immediate processing, minimal storage. Placemeter is building a data layer, not a video storage system. Almost all of the video processed at Placemeter is immediately discarded, and we do not keep a store of historical video about any particular place. We record less than .01% of our videos for quality assurance purposes only.

Continuous improvement. We take our privacy policy and initiatives seriously. We conduct regular internal and independent audits of our privacy practices and work to improve our systems with privacy in mind.

How far can a camera be set up? Could someone use Placemeter to track vehicle movement from a Satellite, for example?

The issue with satellites is that geostationary ones are very far from earth, 36,000 km give or take, so resolution is low. Not enough to see vehicles. For the ones that are closer, resolution is much better – it can down to 10 cm – but they are fast! They will not cover the same area for more than a few seconds. So satellites will not work in short.

But are cameras can be fairly far from their targets nonetheless – we have sensors on top of buildings, or at higher floors in high-rises. As long as your eyes see it, our sensors will see it too.

Who is your target customer?

I’ll emphasize that Placemeter is an open platform that can be used by everyone, for various use cases and in various industries, and this is an important differentiator as competitors focus on only one vertical each.

That said, cities and their governments are a primary customer for us, as are retail businesses large and small. Retailers use us for site selection and understanding foot traffic, which allows them to do interesting things like A/B testing for physical marketing. Cities and local government use us for understanding how people use public spaces, like a park, or how many people visit City Hall.

What are you-have interesting data seen so far?

Placemeter’s goal is to be the platform that makes it easy for users to extract and consume the data they need, which clients and researchers can then use to extract insights for their own qualitative interpretation. A perfect example of this would be Harvard researcher Melissa Sands, who used Placemeter to look at the correlation of 3-1-1 calls and foot traffic in a given area.

 

However, overall, our team has found it interesting that no two places are really alike in terms of pedestrian traffic.

In New York for instance, even streets that are separated by only a few blocks can have vastly different pedestrian patterns. This is an idea that we explore in Blocks of New York, where we profile different blocks around New York, and we intend to eventually look at the unique aspects of block around the world. We’ve found that each block has an individual personality – an individual profile – that is determined by how people interact and travel through it.

Though there are obvious patterns that bubble up, the long-term data tells us just how different a specific place’s foot traffic can be day-to-day. Myriad factors affect pedestrian movement, which you can really only discern with the sort of long-term historical data our urban intelligence platform provides – which emphasizes how inadequate one-week traffic studies can be for someone like an urban designer who wants to design something in a robust manner.

Any plans to release some data for academic research on sociology?

A non-revenue generating audience for us is civic nonprofits, academics, and researchers. We are committed to working with these groups to improve their study and optimize their work. Building better cities is an essential part of our mission and we are proud to partner with these groups in unique ways.

You are coming out with your own sensor this fall, why?

Placemeter’s custom-made sensor, which will be shipped in early Fall, will offer a new option to dramatically increase the ease of deployment and further securing Placemeter’s privacy ­first architecture.

Although our platform is hardware agnostic, there are a few key benefits to our Sensor, which is why we’re excited to offer it as another option for our customers.

When turning video into data, Placemeter uses the same computer vision algorithms to process video whether collected by our Sensor or another camera. However, what is different is when and where the video is processed.

Video is turned into data onboard the Sensor, then only the data is transmitted via internet to the Placemeter platform. The Placemeter Sensor contains a cellular internet capability and can transmit data via cellular internet. IP Cameras transmit the video via the internet, which is then turned into data in the cloud on the Placemeter platform.

The first key benefit for the Placemeter Sensor is not transmitting video over the internet—only transmitting data processed by the onboard algorithms. This means using the Placemeter Sensor requires far less bandwidth than using an IP camera.

The second key benefit is the Placemeter Sensor can transmit data via cellular internet, which alleviates a pain point for customers such as real estate brokers who want to collect data on properties where there is no internet connection.

From a form factor perspective, if you already have several IP cameras and a security system installed in your store or building, you can certainly use those with

the Placemeter platform. However, if you don’t have existing cameras, setting up Placemeter Sensors will be much more user-friendly than installing an IP camera. All you would need to do is place the Sensor on a window overlooking what you want to measure (street, sidewalk, park, etc.). You can adjust the angle of the lens after you install it to get optimal coverage.

The key benefit is saved time and cost compared to installing IP cameras, as well as less maintenance ongoing.

What would you like to see Placemeter that offer technology can not yet offer?

Today we track pedestrian traffic and vehicular traffic alike, including bicycles and motorcycles.

However we treat each vehicle the same today. What we already have in labs and what we expect to release by end of year is an improved algorithm that can differentiate between a car, a bus, a motorcycle, trucks, vans, ambulances, bicycles and more.

In addition to vehicle classifications, this Fall we will roll out new features including but not limited to speed, dwell time in front of a store, and more.

 

Photo by Ömer Ünlü