Peaeater

The Microsoft Live Labs Photosynth Tech Preview is so cool my leg is shaking. It's a pretty fun little application where you fly around in a 3D photospace which has been created from a set of photographs of a particular subject (in the main demo it's Piazza San Marco in Venice).

The interesting bit is how this is achieved. A whole bunch of pictures are taken of a place or object, from all angles, just like you might snap yourself if you were there as a tourist. Then, each picture is examined via funky algorithms for distinctive features and these features are mapped to a photospace so that by the time the analysis is finished, the photospace is a 3D cloud of those features as points in space. Because the photos overlap and share features, they can be automatically fit together in that space to make a three-dimensional collage.

The result, in Photosynth, is that you are able to examine the subject of a collection of photos in a more cohesive and comprehensive way because you understand how each part fits into the whole. At the moment, this is a technology preview, so you can only view the demo collections, but in the future you will be able to do this to your own collections of photographs and relate your own photographs to everyone else's photographs to contribute to a kind of photo gestalt. One of the architects at Microsoft Live Labs, Blaise Aguera y Arcas, describes how the software can join smaller collections together into a larger connected universe in terms of the Wood between the Worlds in C. S. Lewis' Narnia, for which I pretty much fell in love with Blaise Aguera y Arcas (in a completely manly way thank you very much).

The really wild thing is actually not about giving you better tools to inflict your vacation photos on people, but the implications of the "smart photo". The image feature analysis technology at work here can be employed to create image "identities" and to then automatically create associations with similar images and contextual information about the image subject. This concept is immensely powerful: think about how, at present, any correlation between the image of something and a description of that thing must be made inside your head. Imagine being able to take a photo of something to find out more about it.

Some ideas.

• Photospace/map/GPS mashups for waypoint navigation. (Some people are bad at navigating maps top-down and would benefit from seeing landmarks along a route from an on-the-ground perspective.)
• Walking tours where taking a picture of something triggers/fetches a recording or textual description describing it.
• Search engine based on supplied image(s).