Creating and Exploring a Large Photorealistic Virtual Space
Josef Sivic
INRIA - Ecole Normale Superieure, Paris, France.
We present a system for exploring large collections of photos in a
virtual 3D space. Our system does not assume the photographs are of a
single real 3D location, nor that they were taken at the same time.
Instead, we organize the photos according to scene types, such as city
streets or skylines, which we call themes, and let users navigate within
each theme using intuitive 3D controls that include pan, zoom and
rotate. Themes allow us to maintain a coherent semantic meaning to the
tour, while matching on the coarse geometry of the scene allows us to
create a "being there" impression, as if the images were of a particular
location. We present results on a collection of several millions of
images downloaded from Flickr and broken into themes that consist of a
few hundred thousands images each.
Joint work with Biliana Kaneva (MIT), Antonio Torralba (MIT), Shai
Avidan (Adobe) and Bill Freeman (MIT)