Quantifying Reality: People and Objects Turned into Numbers
Enhancing real life with video analytics
What if we could measure reality the way we measure web traffic?
What if we could count anonymous people, track them, see how they
behave, how much time they spend doing something, monitor what draws their attention, observe their behavioral patterns and activity, just like web analytics does? What if we could go beyond that and even identify people’s age and sex?
This idea has led us to create an advanced video analytics system that measures real life.
It analyzes anonymous people’s behavior and movement patterns
in real time: at shopping malls, airports, museums, etc.
The footage from video cameras is then turned into actionable insights, accessible from a user-friendly web panel.
Our system is great for retail, public transport, or tourism.
The gathered data helps to manage waiting lines efficiently, to optimize employee performance, to increase security, or to use product placement effectively.
Perfecting our MovStat algorithms has made it possible for the system to actually learn — once installed, it learns to process video content on its own. Machine learning and AI are our main focuses right now.
Positioning and indoor navigation could be effectuated in other ways,
of course. But Wi-Fi or Bluetooth devices can’t ensure the accuracy and objectivity MovStat can. Not to mention other qualities MovStat has, like analyzing anonymous people’s age.
So far, we’ve conducted pilots at the local airport and in a bank branch.
We measured the passenger flow at security and boarding gates or the average wait time. We analyzed waiting lines and busy hours, we visualized the most popular movement paths in the form of heat maps, and so on…
MovStat also works great in our office. There are 7 video cameras located all over our workspace.
We monitor all the entrances and know how many people come in and out
of the office during the day.
We know how many people go to the kitchen and at what time of the day —
it turns out 12 o’clock is the peak lunch hour. And yes, we do know how much time our employees spend eating their lunch. But relax, it’s anonymous :)
We have a foosball table in our office. MovStat tells us whether someone is currently using it. We also happen to know that the average match lasts 12 minutes :)
We measure how many people enter the conference rooms (and how many
of them choose one room over the other), and so on.
This allows us to effectively redesign our office whenever needed, making
it even more work-friendly.
But the most fun thing regarding MovStat we did was covering
the Microsoft Partner Strategy Conference in Łódź, Poland.
As a Microsoft partner, we decided to enhance the event with
intelligent video content analysis.
The day before the conference we installed 25 MovStat cameras at the venue. Throughout the conference, all day, the cameras relentlessly followed the attendees, analyzing their movement and behavior.
16 video cameras tracking the participants were placed in the main conference hall. Here’a a heat map of the room: the darker the color,
the more crowded the space was. Obviously, the food court was everybody’s favorite area. Interestingly, the attendees preferred the left side of the room (note the orange area in the auditorium: it’s significantly darker than on the other side of the aisle). Did people want to be closer to the food court? :)
Another thing: a lot of conference attendees are late to presentations — this might seem obvious, but we got actual evidence: only 7% of the participants arrived on time. 41% of them were late. 52% were really, really late. Latecomers — we saw that with our cameras! :)
Another video camera was placed near the elevator. 7.04 was the average number of elevator rides per person. 12 o’clock was the rush hour, with 620 people taking the elevator.
Another 2 video cameras were overlooking the street outside the venue during the conference and at the after-party. In total, 5088 cars drove past the venue on that day.
One last (but fun) thing: we video analyzed the after-party as well. Yes,
we know who danced with whom! Just kidding, we don’t, MovStat is all about anonymity and privacy :)
These findings visualize the power of MovStat and numbers. Thanks
to our system, the organizers of the Microsoft Partner Strategy Conference gained valuable insights and can put them to a good use in the future.
MovStat can be the true icing on all conference cakes. Our algorithms provide event organizers, retailers, managers, administrators with accurate data and help them do a better job. We know how to make numbers meaningful
and useful.
