Hao Jiang

Revolutionizing computer vision software

June 1, 2007

Document Tools

Print This Article

E-mail This Page

Font Size
S      M      L      XL

Related Links

By Barry Shell

A computer can watch a video, but it doesn’t know what it’s looking at. "Today’s computer vision systems have a hard time telling whether a fruit is an apple or an orange," says Hao Jiang, one of two SFU 2007 Governor General’s Gold Medal winners. That’s why he devoted the last five years to computer vision research at SFU.

His PhD thesis describes a totally new way for computers to pick out objects in complex settings. In a video of a football game, for instance, Jiang’s software can recognize the quarterback and even the moment he launches a Hail Mary pass. And it can recognize things in two seconds that take conventional vision systems 30 minutes.

Now a postdoctoral fellow at UBC, Jiang is applying this method to video feeds from sports events. The goal is to have computers analyze vast amounts of 2010 Olympic video footage and pick out the highlights.

A native of Harbin, China, Jiang applied to SFU in 2001 on the advice of a visiting professor he met while working at Microsoft Research in Beijing.

Jiang’s mathematical concepts are so radical that his first papers were always rejected. "My original ideas were too foreign to be accepted at first," says Jiang, whose remarkable findings did not come by accident.

"I tried lots of simpler methods that had really bad results, and kept making small improvements that didn’t work either before I tried this radically different approach. We were all surprised when it worked amazingly fast." The technique involves a combination of mathematical ideas that nobody thought could work together. But they do. In fact his technique is now being used for general optimization problems in many areas of study.

Jiang has accepted a position as an assistant professor at Boston College in Chestnut Hill, Mass. beginning this summer.

Search SFU News Online