Quakr XTECH07 17th May, 2007
Is it possible to build a virtual world
from user-contributed photographic metadata?
Quakr Home
This document contains Embedded RDF
Quakr XTECH07 17th May, 2007
XTech 2007, Paris - This talk
- Dave: Laying the Foundations
(The Plan)
- Peter: Bricks and Mortar
(What We Built)
- Katie: What is a "Brick" anyway?
(SWeb Issues)
Part 1
Dave: Laying the Foundations
(The Plan)
The Giants' Approach - Google SketchUp
Google SketchUp
Limitations (1)
Q. Where does Microsoft's 3D data come from?
A. Microsoft
That's a pretty big task to have taken on!
Limitations (2)
Q. How flexible are these worlds?
Arc De Triomphe In The Snow
Arc De Triomphe In The Fog
Arc De Triomphe At Christmas
A. Not very.
Limitations (3)
The installation is taking longer than expected...
An Alternative (1)
Geotagging
- Photographer
- + GPS device
- + Website allowing metadata
- = A geotagged photo
An Alternative (2)
A reasonable question in the year 2000:
"But where are you going to find anyone
even remotely interested in doing this?"
An Alternative (3)
Who saw THAT coming!?
An Alternative (4)
And give the people tools to make this easy...
Flat Mashups (1)
Quakr
A project to build a 3D world from user-contributed photographs
Metadata for photos in 3D (1)
Altitude
Metadata for photos in 3D (2)
Rotation about y-axis
(compass bearing)
Metadata for photos in 3D (3)
Rotation about z-axis
(tilt)
Metadata for photos in 3D (4)
Rotation about x-axis
(roll)
Metadata for photos in 3D (5)
Is it practical for users to measure these rotations?
Metadata for photos in 3D (6)
Metadata for photos in 3D (7)
Quality of the world will improve with:
- increasing user interest
- advancing camera technology
Quakr constraints
Quakr constraints
- Quakr should run in a browser.
- Quakr should require little or (preferably) no installation.
- Quakr should be a thin client, consuming raw resources.
- Quakr should add context information to enhance the 3D experience.
- Quakr should adapt to an ever changing semantic web.
Part 2
Peter: Bricks and Mortar
(What We Built)
Capturing the Data
We need to capture a pile of meta data...
Nokia (550/Sport?) Phone with "tilt" :
Digital Compass and a 3-Axis orientation + compass :
Basically; we want a camera + a GPS device + a "gizmo" :
Homemade - 7D Tiltometer (1)
Basic conclusion - nothing exists... so we built one...
Get a camera and some basic stationery items...
Homemade - 7D Tiltometer (2)
Stick protractors to a camera, add string and plumb bobs :
Homemade - 7D Tiltometer (3)
Add a compass and a GPS device :
Homemade - 7D Tiltometer (4)
One person to take the photos, one to log the readings :
What and how to tag
The choice of tags for the test data was guided by two principles:
- 1. Use the semantically correct term
- 2. Use whatever everyone else is using
Alternative image APIs
Require bounding box query and ability to add tagged meta data
Possibles include;
- http://www.23hq.com

- http://www.photobox.com

- http://www.imagehosting.com

- http://www.yahoo.com/photos

- http://www.google.com/picasa

- http://www.flickr.com

A fully tagged image on Flickr
Machine Tagged...
Flickr API bounding box
Bounding Box API query
"http://api.flickr.com/services/rest/?method=flickr.photos.search
&api_key=XxXxXxXxXxXxXxXxXxXxXxXxXxXxXxXx&min_taken_date=1970-01-01%252000%3A00%3A00
&bbox=-1.3%2C51.754%2C-1.2%2C51.755&tags=quakr&extras=machine_tags"
Flickr API single
A second query for each image
http://farm{farm-id}.static.flickr.com/{server-id}/{id}_{secret}_[mstb].jpg =>
http://farm1.static.flickr.com/164/364490580_426a5c9ec2.jpg
Viewr technology options (1)
Version 0.01: Java applet
Viewr technology options (2)
Version 0.02: VRML/X3D with a plugin
Viewr technology options (3)
Version "Just Works" - Flash + SANDY 3D API
Adding context
Map tiles for floor
Altitude of floor (http://www.earthtools.org)
There are other services...
Part 3
Katie: What is a "Brick" anyway?
(SWeb Issues)
Using Tagging Metadata
"Tagging: describing web content using whatever words seem right" [Beckett06]
- abundant
- recycled
- unhygenic
- evolving
- "emergent taxonomy"?
- "ontological chaos"?
Philosophy of Data Tolerance
- Recycle everything
- Tolerate unhygenic results. (enhance? adjust?)
Benefits of Data tolerance
We, the people..
Costs of Data Tolerance
- Synonyms and Homonyms
- Semantic ambiguity
- Inaccuracy
- Sparseness
- Abundance
- Spam
Synonyms and Homonyms (1)
Synonyms and Homonyms (2)
Synonyms and Homonyms (3)
Semantic Ambiguity One (1)
"Tilt"
Semantic Ambiguity One (2)
What does "Tilt" mean?
Semantic Ambiguity One (3)
What does "Tilt" mean?
Semantic Ambiguity One (4)
What does "Tilt" mean?
Semantic Ambiguity One (5)
What does "Tilt" mean?
Semantic Ambiguity One (6)
What does "Tilt" mean?
Semantic Ambiguity Two (1)
What do "geo:lat" and "geo:long" mean?
Semantic Ambiguity Two (2)
What do "geo:lat" and "geo:long" mean?
Semantic Ambiguity Two (3)
What do "geo:lat" and "geo:long" mean?
Tolerating Sparse Data Sets
Technical approaches:
- Defaulting - tilt
- Enhancement - altitude
Future Work
- Heuristics to handle emerging taxonomies
- Pruning image jungles
- Enhancement of sparse data, using:
- webservices
- co-occuring tags
- More sources - CCTV?
- Geolocated sound?
Conclusion 1 - What we Did
- Aim: 3D world built from community contributed metadata, which runs in a browser
- Geotagging
- What metadata would be needed?
- What are people actually doing?
- Our Flash App
- Which is tolerant and recycles
Conclusion 2 - What we are still struggling with
- Messy tagging data - "heading"/"dir"
- Semantic ambiguity - what does "tilt" mean?
- Inaccuracy - GPS
- Sparse data (enhancement)
- Overabundant data - image jungles
- Spam