Attention Methods

Robert Scoble has a post from November 19 detailing the TechCrunch/Riya launch party. In it, he details a conversation he had with Ojos CEO Munjal Shah. Munjal talked about the facial recognition technology that has us all salivating, but he also revealed something I did not know: that Riya also uses other characteristics to identify certain people in photographs. Scoble’s son was wearing a tie-die shirt that night, and Riya would have stored information about that shirt so that it could better recognize him in case a photo doesn’t have a clear shot of his face.

Now, I’m wondering how well this is going to work in practice. Thousands of people have tie-die shirts, so will it return all those people? I guess it would be cool to say “show me all pictures of people wearing Houston Rocket jerseys” and have Riya return that, but I’m talking more specifically about finding a single person. Do you have to drill down through a bunch of data to find who you are looking for, or does the software match up shirts and other attention details with facial characteristics to give you a closer match?

This is a question I’m hoping to answer whenever the alpha goes out.