🚀 go-pugleaf

                         Universite de Geneve

                           Seminaire du CUI
                                         

 SUJET:  ORGANIZING LARGE STRUCTURAL MODELBASES 
 
 PAR:    Prof. Kim L. Boyer
         Signal Analysis and Machine Perception Laboratory
         Department of Electrical Engineering
         The Ohio State University
         Columbus, Ohio, USA
 
 HEURE:  Jeudi 17 novembre, 1994 a 16h15
         *****
 
 LIEU:   Salle 362, Uni Dufour, 24, rue General-Dufour
               ***
 
 RESUME:
 
 In this talk, we will present a hierarchically structured approach to
 organizing large structural modelbases using an information theoretic
 criterion.  Objects (stored patterns) are modeled in the form of
 random parametric structural descriptions (RPSDs), an extension of the
 parametric structural description graph-theoretic formalism of Boyer
 and Kak. Objects in scenes (observed data) are modeled as parametric
 structural descriptions (PSDs).  The organization process is driven by
 pairwise dissimilarity values between RPSDs.  We also introduce the
 node pointer lists, which are computed offline during the modelbase
 organization.  During recognition, the only computationally
 exponential matching process involved is between the scene PSD and the
 RPSD at the root of the organized tree.  Using the organized hierarchy
 along with the node pointer lists, the remaining work simplifies to a
 series, logarithmic in the number of models, of inexpensive tests,
 each linear in the number of structural primitives (graph nodes), at
 the subsequent levels of the tree search. 
 
 We will develop the basic theory and describe three modelbases we have
 built using these ideas: one of 50 objects built from real image data,
 one of 100 CAD models, and one of 1000 synthetic abstract models.  We
 will also briefly describe a method for growing the modelbase
 incrementally -- avoiding the computational expense of rebuilding the
 hierarchy from "scratch" when a new model is introduced to the set.


________________________________________________________________________
         Pour d'autres informations sur les seminaires CUI,
               veuillez contacter falquet@cui.unige.ch

________________________________________________________________________
 
            Voir aussi par ftp: cui.unige.ch:SEMINAIRES
      ou par WWW: http://cui_www.unige.ch/SEMINAIRES/README.html

________________________________________________________________________