Read/Write World
Blaise Aguera y Arcas
- Background: livelabs, bing maps
- Topics
- photosynth 2008-present
- spatial media: driving/flying, crawling
- reality mashup: putting it together
- future
- Photosynth
- launched in 2008
- Take groups of photos, computer does image processing to stitch together images in 3-dimension.
- demo: 100+ pictures taken of a woman using an ordinary digital camera. you can pan, zoom, rotate around.
- you can see point representation, rotate around an object, zoom in and out.
- it’s a new kind of medium. it’s not a photo, not a video. it’s a 3 dimensional art form.
- you can take pictures of people and bananas, but the really interesting part is buildings and places – photosynths of geographically bound locations.
- demo: bing maps
- zoom into map, street view, etc.
- but also… green zits. these are user contributed photosynths.
- at the metropolitan museum in NY, a user contributed more than 1,000 photos of the greek exhibit.
- it’s enough that the point representation actually shows the entire floor plan, and you can zoom and move around in the space.
- You can get Photosynth for the iPhone…
- Microsoft imagery
- aerial
- oblique
- streetside
- high resolution areal photos of the entire united states and europe. not just cities, but everything. includes height data, terrain mapping, land use maps, building footprints, etc.
- converge
- the meta-data: the images will never all be in the same place, e.g. flickr photos at flickr, map tiles on microsoft servers. but the meta-data, the digital rights, the geospatial data, the semantic connection, named places… all of that is what really need to be converged. RML: real-life markup language.
- service: need a single API to be able to ask “what’s available in this place”? It’s similar to how the value of the web is unlocked by having a search engine that indexes the web.
- viewer:
- do it once, do it right…
- read/write
- realtime
- nosql
- uniformly represented rights
- opensource viewers
- mashup-friendly
- consider…
- keyframe + tracking = strong augmented reality.
- weak augmented reality is using the GPS and compass to approximate where things are.
- but strong augmented reality is taking the photo, matching it against the database, and now doing a highly precise annotation of the visual field.
- questions
- Q: What about the privacy issues?
- every country has different rules
- the stuff that is captured by microsoft is only in the public view, what anyone can see walking by…
- the highest value imagery is downtown, businesses, stuff where people want visibility.
- the privacy implications of indexing people’s images… it’s like the implications of indexing people’s web pages. in the beginning, there were people who put stuff up on the web, and figured it would be private if they just didn’t tell anyone the url. indexing search engines made this an unworkable approach to privacy, but added a ton of value to the web, and other tools (e.g. facebook) arose to maintain privacy where it is desired.
- Q: temporal aspects to this – building facades over time, a construction site.
- Yes, you want to be able to time travel as well as space travel. We have archival stuff no, we just haven’t added explicit time navigation.