I wrote a couple of posts last year about the expanding field of companies offering content-recognition services to user-generated video sites, peer-to-peer networks and other businesses with inventories of uncertain provenance. At the DEMO conference this week, yet another firm joined the fray: Eyealike, a small company from Bellevue, WA, whose strength is in facial recognition technology.
Eyealike has been in talks with social/dating sites to enable searching by image (e.g., letting you upload a picture of your ex so you can date his/her doppelgänger). In the meantime, President Greg Heuss said, the company has been extending its capabilities to recognize moving images. Its Eyealike Copyright technology examines each frame for faces or other recognizable objects, along with patterns of motion, color and other characteristics, Heuss said in a recent interview. Among other intriguing applications, Heuss said the technology can identify objects composited into a frame (as in video mash-ups), as well as videos shown on screens within the frame (i.e., picture-in-picture).
According to Heuss, the company plans to start offering Eyealike Copyright later this year but is still exploring ways to monetize it. The major entertainment companies he's spoken to are divided internally into factions over the use of content-recognition technology: lawyers who want to stop copyrighted material from appearing online, and marketers who'd rather collect revenue when clips are posted and played by fans. The latter approach has been hindered, Heuss said, because it's not clear how much sites such as YouTube should pay when copyrighted clips are viewed. Content-recognition technology can help those negotiations, he added, by revealing just how often such clips get played.
Although Eyealike is well behind the first movers in the market, the content-recognition field is still in its infancy. Companies with brands well known in Hollywood, such as Nielsen, are just getting in, as are European broadcast-industry vendors such as iPharro. Nevertheless, companies aren't likely to succeed unless their technology can identify at eye-blink speed an immense number of unknown files. Otherwise, they won't be able to meet the needs of popular user-generated sites and file-sharing networks. They'll also have to persuade the major studios to help them build comprehensive databases of audio and/or video fingerprints, which they'll use to identify the files they find online. It's hard to image the studios supporting multiple content-ID databases, so it's a safe bet that the field will consolidate significantly before too long.