Google acquires facial-recognition company PittPatt

Background

Facial recognition is a branch of computer vision (the field of getting software to interpret images). A system first locates a face in a picture, then turns the arrangement of features into a numerical description that can be compared with descriptions taken from other pictures. Done well, this lets software answer a question that people answer instantly: who is this, and where else does this person appear?

Google already held very large collections of exactly the material such a system feeds on — photographs and video uploaded by users.

What happened

In July 2011, ZDNET reported that Google had acquired Pittsburgh Pattern Recognition, better known as PittPatt, bringing the company's facial recognition technology into Google ^[1]. A review of Google's artificial-intelligence acquisitions published by Jiemian dates the deal to 23 July 2011 ^[3]. According to Computerworld, the terms of the deal were not disclosed ^[2].

The company behind the technology

PittPatt's founders began developing the technology at Carnegie Mellon University's Robotics Institute in the 1990s, and the company itself was formed in 2004 as a spin-off from the university, Computerworld reported ^[2].

The same report described what the software could do: it could track faces in videos, and it could sort photos according to whose face appeared in each photo ^[2]. Both tasks are harder than they sound. Tracking a face across frames of video means coping with changing angles, lighting and expression while keeping hold of the same identity; sorting a photo collection means comparing every face found in every image against every other one.

How it was meant to fit into Google

PittPatt said its approach was a natural fit for Google products such as Image Search, Picasa, YouTube and Google Goggles ^[1]. Those products cover searching for images, storing and organising personal photos, hosting video, and visual search from a phone camera — in each case, software that can tell one face from another would have something to work with.

A Google spokesperson said the PittPatt team had developed innovative technology in pattern recognition and computer vision ^[1].

Why it mattered

The acquisition is a small but telling step in how Google built up its capabilities in artificial intelligence: rather than only hiring researchers, it bought small companies that had already spent years on a narrow technical problem. Jiemian's review of Google's AI acquisitions places the PittPatt deal within that longer series of purchases ^[3].

For an ordinary user, the practical stake is organisation: a photo library or a video archive becomes searchable and sortable by person once the software can reliably pick faces out of it ^[2]. That is also why the technology attracts attention beyond the engineering — the same ability that sorts a holiday album can, in other hands, be used to identify people without their cooperation.

The terms of the PittPatt deal were never made public, so what Google paid, and how much of the technology later appeared inside shipping products, was not part of the record at the time of the acquisition ^[2].