Snavely is working on computer graphics research in an effort to stitch together a digital 3D rendering of entire monuments, cities, and potentially the world.
“Broadly speaking,” Snavely says, “I’m working on reconstructing the world in 3D from photos on the Internet.” His research is developing techniques to calibrate a “world camera” utilizing the multitude of digital photo data available on the Internet.
He is working to stitch together a representation of the world from the data that already exists and is created daily by people around the world. Snavely suggests, “This visual data on the Web represents a massive, distributed camera, made up of millions of sensors around the world.”
Juxtapositions of photos from online sharing communities such as Flickr and PhotoCity - a site that Snavely launched which challenges Cornell and University of Washington to build 3D models of their campuses - are used to create renderings of various places. “By comparing these photos to each other, our techniques can reason about the 3D geometry of different photo collections, automatically figuring out where each photo was taken,” explains Snavely.
The project has recreated monuments such as the Colosseum, Trevi Fountain, and St. Peter’s Bascillica. They face the challenge of not only processing hundreds of images from many different angles, but also the issue of optimizing said processing. Snavely is practiced in this field of computer graphics and tools. He has contributed photo recognition research, reconstruction research, and co-authored a paper which won runner-up in the 2011 Computer Vision and Patter Recognition Conference.
The research, according to Snavely, will be useful in many fields, from tourism to farm management. “Imagine monitoring crop growth, measuring differences in the onset of spring in different places each year, or going back to the photo record to do a kind of forensics –– trying to figure out why a bridge collapsed, for instance –– all from massive collections of photos on the web.” The research will work to preserve data which can be used to depict changes over time or as a tool for educational or promotional purposes.
“I was inspired to pursue this work by the amazing quantity of imagery available today –– the billions and billions of images on the Web –– and the difficulty in digesting ... all of this visual data,” reflects Snavely. He plans to utilize his Microsoft Fellowship funds in order to expand his research group.