COMMENT ⊗   VALID 00010 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	Some other names for Pajaro Dunes
C00003 00003	make films
C00005 00004	LOU,RCB,WAP,BGB,MJH,PDQ,KKP
C00007 00005	Categorize by goals (how many, how complex)
C00010 00006			VISUAL PERCEPTION AND AUTOMATION
C00014 00007	using our representation of complex objects, describe a few objects for
C00015 00008
C00016 00009	Informal followup discussions later in week
C00017 00010	
C00018 ENDMK
C⊗;
Some other names for Pajaro Dunes

Harry Andrews--USC
Wm Pratt--USC
Bob McGhee--Ohio State
Sam Dwyer--Univ of Missouri
Charles Harlow--Univ of Missouri, Columbia, Mo
Prof. Riseman--Univ of Mass, Amherst
Bruce McCormick--Univ of Illinois,Chicago
Jack Sklansky--UC, Irvine

make films
	disable timeout on III controller and work lightly loaded
	other ways
		computer controlled single frame
		to run dpy n times
			FOO,,74;RESTORE FLAGS
			<BYTE(8)0,0,200>+12;TEST CONTROL BIT
			-1,,20;SHUTS OFF WORLD
			<BYTE (8)0,40,0>+12;ZERO EDGE OVERFLOW AND DON'T SKIP
			<BYTE(11)DX,0>+46;MOVE RIGHT DX
			<BYTE(8)0,0,40>+12;TEST EDGE OVFL, SKIP IF NOT
			<BYTE(8)200>+12;IF NO SKIP,TURN ON CONTROL BIT
			0
			FOO,,64	;SAVE FLAGS


film prints
	slight advantage at 4 copies for internegative and optical track
	.081$/ft 1-9 copies from internegative
	.34$/ft for internegative and optical track
	vs .18$/ft electroprint from A/B rolls and consumes original slightly
	
		DX=1024/N
			INIT FOO TO 0


protection usable with alias
	RENAME ←FOO.BAR/P=205

operate at MIT
	:? gets list of commands
	:?? gets longer list of commands
	:LISTF GJS;	FILE DIRECTORY
	:PRINT		TYPES OUT FILE

SAIL DEBUG
	RU SAILDD[S,AIL]
	*←FLOWER/1M

	L3600/7
	2M#P
	1M#P
	3M#P

LOU,RCB,WAP,BGB,MJH,PDQ,KKP
BGB DEMOS

RUN 0[1,BGB]	! arm motion display;
	n	! any number slows down action;

R G			! geomed;
	ctrl ∞		! make torus;
	alt-mode	! hidden line;
	H		! help;


R CRE
	ctrl I	HIP[CRE,BGB]	! input, file name for horse;
		DOL[CRE,BGB]
	meta ↔		! play film of images;
	H		! histograms;
	Q		! contours at 20,40,60;
HELP CRE		! at monitor level;


AL GEM,BGB


RU G
	I	HIP0		! horse;
	ctrl I	TMP		! camera model for horse;
	meta Z	HEAD
	meta Z	BASE


R G
	I	DODECA[B3D,BGB]		! dodecahedron;
	I	STAR[B3D,BGB]
	ctrl¬		! subtraction of bodies;
	ctrl ∪		! union of bodies;
	ctrl ∩		! intersection of bodies;
	K 		! kill;
	C		! copy;

Categorize by goals (how many, how complex)
  by objects(how many, how complex)
  by knowledge available, required

We discuss these issues in the context of some paradigms:
PARADIGM 1:  We know what is in the visual environment, where the
objects are, and remember colors and details of all objects.  Moreover,
we know that nothing else is in the environment.
PARADIGM 2:  We know some of the objects, and some loose spatial
relations among them.  We do not know what else is in the environment.
PARADIGM 3:  We are in a "known" environment, such as an ordinary office,
or outside, with a complex set of goals, complex environment and
rich universe of objects.
PARADIGM 4: We are shown a photograph without knowing the context.

In the first paradigm, if the goals are few and objects few,
some relatively simple techniques may be used to recognize objects
[Tenenbaum, Freuder].  A tlephone may be the only black object.  We
note that Tenenbaum's work uses no shape information, only pointwise
properties such as color, and depth from a ranging device.

HOW WILL THESE GENERALIZE

FUNDAMENTAL VISION PROBLEMS
segmentation-body,part
obscuration-indexing
representation
computational complexity
linking-indexing
description vs recognition
use of context
learning
low level,bottom up  vs high level,top down

How does each appear
What are the options
		VISUAL PERCEPTION AND AUTOMATION


Minsky and Papert argue that vision is a prime problem area for
the study of intelligence.  They see the main problems as the description
of data and procedures to represent a variety of kinds of knowledge.  This
knowledge we now conceive of in terms of microtheories, well worked out
systems of knowledge in small domains such as topology, geometry, chemistry,
biology, psychology and human relations.  We come to see the problem as
implementing these microtheories as the semantics of particular domains.
The most important microtheories of vision are related to representations.
Two representations help us with problems like:
	how do we describe a complex object
	how do we relate a scene just after moving our head to that
		just before.
These are the representations of shape and space.  As problem areas,
the microtheories of vision are relatively well-defined, since they
are derived from models of topology and geometry, and properties of the
physical world.  Some of the complication of social and psychological models
does not widely enter in vision.  

Some areas which have been attacked have been:
	theoretical understanding of line-drawings  Huffman,Clowes,Waltz
	structural learning	Winston
	use of world model for outdoor scenes	Yakimovsky
	representation of complex objects	Binford,Agin,Nevatia
	computational geometry-computational complexity
There is a great need and opportunity for work in perceptual strategies
which has the advantage of a well-defined problem domain for which we
have great intuition, so that we can use semantics to do otherwise very
hard problems.  There is also a need for good systems work.

This is an applications area with promise.  We have a serious effort in
programmable automation.  The application area is a source of good problems
in AI at this stage.  We hope that more support for AI will come from the
realization of the impact of these techniques on problems outside the
university.
using our representation of complex objects, describe a few objects for
	example, pump base

extend the ideas of Huffman and Clowes with dual graphs (see Mackworth))

fourier transform edge verifier

Texture analysis:

	2-d shape description by projection
	cluster descriptors

improve convergence of calibration model by choosing improved non-singular
	parameters

stereo correlation, analysis and programming

edge finder for curves

use visual feedback to put a screw into a hole.

















				VISION SEMINAR







				    TOPIC
	 	   ANALYSIS OF LINE DRAWINGS AND SURFACES
		   A Lecture Series by Prof David Huffman












				FIRST MEETING
			    3PM, 14 JANUARY,1974
			    Serra House Conf Room


		    Realizable and Unrealizable Objects
			   Prof David Huffman
Informal followup discussions later in week

Lectures as planned
2.  Dual picture and some of its applications
3.  Realizability of Pictures of General Polyhedra
4.  Change of Vantage Point and 
	Determination of Illumination and Shade
5.  Non-Planar surface; Intersections and Shadows
6.  Zero Curvature Surfaces
7.  Paper Folding