On the streets of Pompeii lie some of the earliest crosswalks in human history. Raised stone blocks once offered safe passage for pedestrians crossing roads that doubled as sewage and drainage systems.

Nearly two thousand years later, the rise of car traffic presents a different problem for those navigating streets on foot. Despite a plethora of crosswalk designs, crossing the street can still be a perilous endeavour.

For families under the California sun, beyond the gates of the happiest place on earth, the challenges are not about road safety, but about user experience. Here, long line-ups, crowded paths and vast spans of park present obstacles for families taking in the Disneyland experience – and their desire to return.

In the business of safety or the business of entertainment, gathering data about human behaviour used to be prohibitively time consuming and inaccurate. Today, the prevalence of surveillance video has largely solved the data gathering problem, but it has also created new ones: volume and velocity. Locked in the timecodes of millions of hours of video footage lies the answers to safer environments, and happier customers. But how do we unlock the data in motion and extract meaningful information — quickly and efficiently?

This is the conundrum that public and private sector industry across North America bring to Simon Fraser University professor Greg Mori. From a modest lab in the corner of the applied sciences building, Mori and his team are leading the field of computer vision and machine learning to improve human experiences, operations and decision-making through data. Empowered by new scales of real world video data and powerful computational approaches, his work is measuring aspects of human behaviour that used to be difficult to measure.

 

“In computer vision, we are developing algorithms that can automatically understand images and videos,” he explains. “In today's world cameras are ubiquitous and they're collecting both video data and imagery pretty much around the clock and having algorithms that can help us make sense of this data is very important.”

It’s contrary to common notions of “big data” — spreadsheets, web analytics, financials or outcomes of quantitative research studies. Computer vision automates data extraction through machine learning. Initially, humans label a certain amount of data themselves and then use that data to train an algorithm to extract more data. For an individual, this can be useful for organizing a personal video collection. For a company like Google, it is useful for placing the right ads with the right videos on YouTube.

Using computers to study movement repeatedly over time can identify and help solve limitless problems for businesses and the public sector – from reducing the number of seniors falling in long-term care homes, to surveilling crosswalks to assess what’s more effective: pedestrian crossing signs, a zebra striped pattern, flashing lights or a pedestrian controlled stop light. 

“Computer science is exciting because it's intellectually stimulating and you can have a big impact on society. If you're able to write software that can analyze data, you have this superpower that can let you change the world."

“The scientific questions we try to answer are: how do people tend to behave in a particular environment? What types of interactions do they have both with objects in their environment and with other people, and can we develop algorithms that can automatically understand those questions?” he explains.

Mori works with a long and diverse list of industry partners, from startups like Sportlogiq to global brands such as Disney and Google, who benefit from the unique nature of working with researchers in the academy.

“The companies we partner with receive advice and consulting from myself or other faculty members who let them know what state-of-the-art algorithms exist, and how might those algorithms perform on their data,” he explains. “We're going to be able to answer more fundamental questions, develop longer term approaches to solve problems and develop riskier prototypes that an industry partner may not have the resources nor the personnel to explore otherwise.”

“In computer vision, we are developing algorithms that can automatically understand images and videos. In today's world cameras are ubiquitous and they're collecting both video data and imagery pretty much around the clock and having algorithms that can help us make sense of this data is very important.”

Partners also benefit from access to talent in SFU’s Professional Masters in Big Data program. While the scale of big data continues to grow exponentially, so does the demand for job-ready leaders in the field, equipped with capabilities and industry experience to hit the ground running. For the 40 students who will graduate the program this year, Mori’s partnerships are a boon to their career prospects.

“There are many people who either need big data analytics or people to help them with building software and developing algorithms to analyze data,” Mori explains. “The demand is measured in the hundreds of thousands. But there are maybe only a few thousand students being trained in the world who do this. Our 40 graduates are going to do very well in the job market.”

But Mori also sees the opportunities that lie ahead of SFU computer science grads, beyond the high paying jobs with tech startups and industry giants.

“Computer science is exciting because it's intellectually stimulating and you can have a big impact on society,” he says. “You can convert whatever thing you care about in the world, whether it's sports analytics or whether it's designing safer road systems or studying environmental impacts of some chemical. If you're able to write software that can analyze data, you have this superpower that can let you change the world.”

Each time Mori stops in to chat with students in his lab, he is not only teaching a new generation this “superpower” — he is also inspiring them to use their talents to make the world a better, safer place, frame by frame, street by street.

Sign up to learn more about big data at SFU

Sign up