70 likes | 164 Views
Tool for pictorial contents retrieval. IST KA3: Semantic Web Technologies Workshop. department of computer science and engineering. Zdenek Mikovec, Pavel Slavik Czech Technical University Prague Department of Computer Science and Engineering. November 2000. Content.
E N D
Tool for pictorial contents retrieval IST KA3: Semantic Web Technologies Workshop department of computer science and engineering Zdenek Mikovec, Pavel Slavik Czech Technical University Prague Department of Computer Science and Engineering November 2000
Content • motivation, problems & solutions • implementation • results and future work
Motivation, problems&solutions • traditional approach to picture information retrieval: histograms, keywords, ... • problems & solutions • P: pure description of picture structure&semantic • S: introduction of relations among picture objects • relations could be of two principal classes: • structural (they reflect visual structure) • functional (they reflect semantic relations among object) • P: efficient information retrieval from pictures • S: information filtering and pictorial information inference • P: query result presentation in a suitable picture format • S: suitable formal description of transformation process
Implementation • all mentioned requirements could be satisfied by the use of XML and XSL transformation sheets 2a. 2b. 3. new relation specification (XSL) filtering information (XQL) spatial specification (XSL) DBS (XML) DBS (XML) DBS (XML) DBS (XML) import (XSL) visualization (XSL) requests metric definition 4. 1. MPEG-7 WebCGM (XML) VRML Web CGM MPEG-7 feedback description browsers 5. BROWSING USER XSL pipeline
DBS DBS DBS DBS new rel. generation filtering inform spatial specification requests import metric definition visualiza-tion choose import sheet choose visualization sheet MPEG-7 WebCGM Web CGM MPEG-7 VRML feedback BROWSING USER Results and future work • described system • semantic views and filters • simpler querying • corresponds with current trends (MPEG-7) • system has been implemented and tested • picture databases on WAP • picture information retrieval for blind people Pre-defined dynamically adapting XSL pipeline
Information inference <?XML version="1.0"?> <movie name="M.A.S.H."> <popularity>7</popularity> <type>comedy</type> <funny>7</funny> </movie> <movie name="Indiana Jones" <popularity>8</popularity> <type>action</type> <funny>4</funny> </movie> <?XML version="1.0"?> <movie name="M.A.S.H."> <popularity>7</popularity> <type>comedy</type> </movie> <movie name="Indiana Jones" <popularity>9</popularity> <type>action</type> </movie> trans- formation Original DBS (XML) DBS with new relations (XML) type.comedy=1 type.action=0.5 funny=type*popularity New information - transformation rules (XSL)