JHC%OZ.AI.MIT.EDU@XX.LCS.MIT.EDU (03/18/86)
[Forwarded from the MIT bboard by SAWS@MC.LCS.MIT.EDU.]
Thursday , March 20 4:00pm Room: NE43- 8th floor Playroom
The Artificial Intelligence Lab
Revolving Seminar Series
LEARNING SYMBOLIC OBJECT MODELS FROM IMAGES
Jonathan Connell
AI Lab, MIT
This talk will present the results of an implemented system for
learning structural prototypes of objects directly from gray-scale
images. The vision component of this system employs Brady's Smoothed
Local Symmetries to divide an object into parts which are then
described symbolically. The learning component takes these
descriptions and forms a model of the examples presented in a manner
similar to Winston's ANALOGY program. The problem of matching complex
structured descriptions and the difficult task of reasoning about
function from form will also be briefly discussed.
Refreshments at 3:30