How many people turn up for an event is in itself newsworthy, even more so if it's a protest march. The pro-democracy protest on Monday in Hong Kong is a prime example, according to The New York Times the organisers claimed that 550,000 people turned out, the police counter-claimed with 190,000, but AI researchers put the number at 265,000.
When the number of people you are counting surpasses 100,000 then you have probably gone past the traditional method of having people stationed along the route manually counting individual participants. And while AI "doesn't make the calculation definitive" the authors of the report claim that "the technology helps produce a more precise estimate because it uses computers to try to count every person."
Here's the explanation about how AI technology was applied to counting the Hong Kong protest earlier this week:
"On Monday, the A.I. team attached seven iPads to two major footbridges along the march route. Volunteers doing manual counts were also stationed next to the cameras, to help verify the computer count," the article explains.
"The team's software ran multiple models with different parameters to track people in the crowd. People could be counted when they crossed a so-called counting line within the frame of the video. Throughout the day, the team monitored the results and adapted to changing variables like the density of people, the speed of flow and different lighting conditions."
AI isn't the only way of using technology to count crowds, an article in Wired explains how 'heat mapping' using mobile phone or wifi signals can be used to estimate crowd numbers based on the amount of devices emitting a signal. This has its downsides in that people not participating in the protest - that is, those who might be inside a shop along the route - will also be counted. Also, in a huge crowd mobile phone networks can quickly become overloaded and unreliable.
The Wired article was referring to the anti-Brexit march in Britain in March this year, in which some claimed a million people took part, while official estimates were that less than 400,000 turned out.
It seems that whatever technology you opt for, the bias is to overestimate the size of the crowd. That's because as human's we have tendency to overestimate everything, as one of the experts quoted in the article points out: "Ask me how long I was on-hold on the phone the other day. I'd probably say five to ten minutes. My phone log would say one to two minutes."
You must be logged in in order to post comments. Log In