Top of Page

A face in the crowd: Pittsburgh Pattern Recognition can find it

on
Finding a face in the crowd–specifically, in the mobs of digital video images being uploaded to the Web or on security cameras–has become easier of late, thanks to the work of Pittsburgh Pattern Recognition.

The five-year old spinoff from Carnegie Mellon University has just won additional funding to assemble its face detection, face tracking, and face recognition software for turnkey operations for its customers.

At its home in Pittsburgh’s Strip District, the firm has created a software development kit (SDK) that provides access to function to finding faces, tracking, and matching one face to another. “It works very well in simple applications,” says Mike Sipe, its VP for product development. “ For example, a Webcam identifies a face instead of a password. We want to expand and more fully exploit emerging market for video analysis, more of a turnkey solution.”

Internet video, security video, professionally produced entertainment video, amateur video, video conferencing are all growing at an explosive rate, and it’s human faces that the primary objects of interest, says Sipe.

“Content producers and advertisers have a problem. To attract an audience, you need to be able to find (a video subject). If you want to find a segment of video–say, Kareem Abdul Jabar with the Lakers–you don’t want to look at whole basketball games, you want to find scenes with him.”

PittPatt grabbed the Web spotlight with a hugely popular demo of its software that recognized and analyzed Star Trek characters over hundreds of episodes. Sipe says the face mining demonstration generated “an absolute flood of inquiries and hits” that prove the company’s technology is right on the nose.

Source: Mike Sipe, Pittsburgh Pattern Recognition
Writer: Chris O’Toole

Features, News
Top