In one of Agüera y Arcas' previous projects, Photosynth, his team tried to model the world from tourists' photographs. They came to realize it would require a framework or "trellis," upon which to organize these fruits. In this talk, Agüera y Arcas explains the resultant structure of Bing Maps. Bing Maps is a “refresh” of Virtual Earth, more tightly integrated with Bing. It provides a spatial canvas with rich semantic data bound to it – the goal remains to have a model of the earth's surface, capable of continuous update, which can be used as a data resource for apps and services riding on top.
Bing benefits from several realizations about image search. One was considering screen resolution as a limiting factor in graphics load time, and reorganizing image data to suit. Two was to use processing time for analysis of uploading photos, the kind of analysis that could determine point-of-view well enough to map data clouds from 2D images into 3D models, and generate a continuous axonometric representation from ground and aerial photographs. The third was to make realizations generalizing the subjects people photograph. The fourth is to give users the tools so that they can help in the organization of images into a model of the world. Once users were given the opportunity to upload bound sets of images, they were able to infill information in the data clouds while directing their own 3D reconstructions onsite.
Microsoft's Silverlight analyzes the semantic content of photographs, such that they can be compared and connected to others. In important tourist sites, there is enough data to create a 3D image of these places from data cloud information from sets of 2D images. Agüera y Arcas notes that geo-registered photographs can still be off as much as 40 meters; so this information has a limited use in recreating 3D from data clouds.
Blaise Agüera y Arcas is the Architect of Bing Maps and MSN at Microsoft. He leads an Advanced Engineering team of researchers and engineers with strengths in social media, computer vision, and graphics. His startup company, Seadragon, was acquired by Live Labs in 2006. In 2007, he presented one of the most talked-about TED Talks, on Photosynth. Blaise has worked in computational neuroscience, computational drug design, and data compression. In 2001 he received press coverage for his discovery, using computational methods, of the printing technology used by Johann Gutenberg. Blaise’s work on early printing was the subject of a BBC Open University documentary, entitled “What Did Gutenberg Invent?”. He has published essays and research papers in theoretical biology, neuroscience, and history in The EMBO Journal, Neural Computation, and Nature. In 2008-9 he was a recipient of MIT Technology Review’s TR35 award and Fast Company’s MCP100.
Resources
This free podcast is from our Where Conference series.
For The Conversations Network: