🚀 go-pugleaf

           Department of Computer Science, University of Toronto
             (GB = Gailbraith Building, 35 St. George Street)

       -------------------------------------------------------------

                      ARTIFICIAL INTELLIGENCE SEMINAR
              GB305, at 11:00 a.m., Thursday 22 February 1990

                            Dr. Katushi Ikeuchi
                        Carnegie Mellon University

            "Vision Algorithm Compiler for Object Localization"

This talk describes a vision algorithm compiler, which automatically
generates an object localization program from a given geometric model of an
object.  The compiler utilizes aspects, topologically equivalent classes of
object appearances, as a basic tool for object localization.  The compiler
will embody the following two-step strategy into the localization program:
first to classify one unknown appearance into one of the aspects and then
to determine the precise attitude within the aspect determined.  We will
discuss how to:

extract possible aspects,

describe aspects,

generate aspect classification parts, and

generate attitude determination parts.

We will generate several localization programs by using the compiler from
several object models, and then apply the generated programs to real scenes
to demonstrate validity of our vision algorithm compiler.