View
179
Download
2
Category
Preview:
DESCRIPTION
ABSTRACT: Creating and maintaining scenes for mobile Augmented Reality browsers can be a challenging and time consuming task. The timeliness of digital information artifacts connected to changing urban environments require authors to constantly update the structural representations of augmented scenes or to accept that the information will soon be outdated. We investigated approaches for retrieving multimedia content and relevant web services for mobile Augmented Reality applications at runtime. Using semantic web technologies we are able to postpone the retrieval of actual media items to the moment a user actually perceives an augmented scene. This allows content creators to augment a scene only once and avoid continous manual updates.
Citation preview
Jens Grubert+, Lyndon Nixon†, Gerhard Reitmayr+
+Institute for Computer Graphics and Vision†STI International
Augmenting the World using Semantic Web Technologies
Do I have time?
Are my friends attending?
What kind of music they play?
How much are the tickets?
Do I have time?
Are my friends attending?
What kind of music they play?
How much are the tickets?
As time goes by …
How to create information?
How to update information?
Standard approach:direct links to content
Creatingaugmented media
…<mainresource>http://www.klimmstein.com/before.jpg</mainresource>…
Standard approach:manual update
…<mainresource>http://www.klimmstein.com/before.jpg</mainresource>…
outdated information
after.jpg
Updatingaugmented media
Link to concepts, not to content
Retrieve the content at runtime
Our approach
Less effort for updating augmented scenes
Potential added value due to context-sensitive retrieval
Link to concepts, not to content
Retrieve the content at runtime
Our approach
ConceptLink to structured machine-readable data in
ontologies (Linked Data)
Link to concepts, not to content
• Events: o Clubbings, festivalso Small to midscale
• Authors: o Event organizerso public relations
Link to concepts, not to content
Link to concepts, not to content
concept (URI or tag)
display region
Link to concepts, not to content
TOI reposit
ory
TOIThing of Interest
structured description of an augmented object
<toi:segment_annotation …><toi:representation> <toi:hasDisplayLocation>0.59, 0.03</toi:hasDisplayLocation> <toi:hasDisplayDimension>0.44, 0.31</toi:hasDisplayDimension> <toi:hasLocatingInformation>0.11, 0.03</toi:hasLocatingInformation> <toi:hasDimension>0.44, 0.34</toi:hasDimension></toi:representation> <toi:represents> <toi:identifier="http://data.play.fm/artist/chrisharvey"> <toi:classification rdf:resource="http://www.play.fm/ontology#Artist"/></toi:represents></toi:segment_annotation>
Link to concepts, not to content
Retrieve the content at runtime
Our approach
TOI Reposit
ory
SmartReality Server
Web of Data
Semantic Web Services
Augmented Reality Client
Object Recognition
Thing-of-InterestInitialization
Content Retrieval
Tracking and Rendering
Image Recognition
DB
Retrieve the content at runtime
Image Recognition
DB
ID
Millions of Images
Augmented Reality Client
Object Recognition
TOI Reposit
ory
SmartReality Server
Augmented Reality Client
Thing-of-InterestInitializationImage ID
Sesame Triple Store
http://www.smartreality.at/rdf/toi
How is it represented?
What does it represent?
(classification)
Thing-of-Interest (TOI) TOITOITOI…
Linking Open Data cloud diagram, Sept. 2011. http://lod-cloud.net/
Artist, Event, Club, Show, Recording, Track
Timeline, Event, Dublin Core
Uses:
SmartReality Server
Augmented Reality Client
Content Retrieval
Linked Open Data Collector
TOIclassification
Crawling Rules RDF/N3
e.g., for each Music Group crawl also each
member
+BBC Music
MusicBrainz
SmartReality Server
Semantic Web Services
Augmented Reality Client
Content Retrieval
RDF Service Repository
(WSMO-Lite) Discover relevant web services
Rank and execute web services
How are the Posters annotated?
who | what?
when?
where?
• Authors: o event organizerso PR personal
• Events: o clubbings, festivalso small to mid scale
<toi:segment_annotation …><toi:representation> <toi:hasDisplayLocation>0.59, 0.03</toi:hasDisplayLocation> <toi:hasDisplayDimension>0.44, 0.31</toi:hasDisplayDimension> <toi:hasLocatingInformation>0.11, 0.03</toi:hasLocatingInformation> <toi:hasDimension>0.44, 0.34</toi:hasDimension></toi:representation> <toi:represents> <toi:identifier="http://data.play.fm/artist/chrisharvey"> <toi:classification rdf:resource="http://www.play.fm/ontology#Artist"/></toi:represents></toi:segment_annotation>
Challenges
• Effort to extend it beyond existing use case?o Suitable Ontologies o Crawling Rules
• Semantic Descriptions of 3D Models?
• Performance (roundtrip currently ~ 10 seconds)
• How to capture the context?o User Model, Context Model (e.g., FOAF, Activity Streams)
Summary
• Goalso Context-aware information access ion real objects in-situo Minimize creation effort for authors (non-programmers)
• How?o Semantic Web Technologies for content retrievalo Augmented Reality for content presentation
• Challengeso Not yet able to model visually compelling (3D) scenes
This work is made possible by the Austrian National Research Funding Agency FFG in the
SmartReality project.
Thank you for your attention.
www.smartreality.at
Recommended