New Scientist Technology - Camera phone helps label snaps

New Scientist has a story this month on some of Marc Davis's work at the new Yahoo Labs at Berkeley where they are exploring ways of further automating the capture and intelligent use of metadata from cameraphone pictures. Some of the inference approaches are particularly interesting.

The project is called Mobile Media Metadata 2 (MMM2) and is an extension of work done in the Garage Cinema Research lab. One of the unique approaches used in MMM2 is "bluetooth-sensed human co-presence" (remember, these are researchers!). What this means is that if the system knows where your phone is from cell-tower information from the network (or GPS in the phone in the future) and it knows the location of other phones close to your phone using Bluetooth (or maybe it can figure that out from cell tower proximity), then this information along with additional contextual information could be used in some interesting ways.

I could be used to drive "share guesser" which gives you suggestions on who you might want to share a particular photo with (this also uses past sharing patterns). I can also be a useful variable to be combined with things like face recognition. You could see how services like Riya could use this to help identify people in your pictures.

The Yahoo/Berkeley partnership is an interesting one since it is leveraging an academic setting rather than trying to export researchers into a corporate lab. It will be interesting to see if and how Yahoo makes use of some of these innovations.