perm filename CORRES[DOC,TOB]8 blob
sn#165637 filedate 1975-06-24 generic text, type C, neo UTF8
COMMENT ā VALID 00012 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 Ganapathy
C00008 00003 request for Turner thesis
C00012 00004 Yasuda
C00018 00005 Thomas
C00022 00006 Ganapathy
C00026 00007 Ganapathy
C00030 00008 Ganapathy
C00033 00009 Miyamoto, April 16,1975
C00035 00010 John McCarthy April 16,1975
C00042 00011 Kidode April 7,1975
C00044 00012 John McCarthy April 7,1975
C00050 ENDMK
Cā;
Ganapathy
Stanford Artificial Intelligence Laboratory
Computer Science Dept
Stanford University
Stanford, California,94305
Prof Aaron Sloman
Cognitive Studies Programme
School of Social Sciences
The University of Sussex
Brighton, BN1 9QN
England
Dear Prof Sloman:
I am writing to support Sundaram (Kicha) Ganapathy's application for
a position in your group. He has specialized in computer vision and
has some related research interests as well. He has shown strong
interests in theoretical problems in combinatorics, and produced some
results, in addition to his work in computer vision. He has done
considerable programming, although that has not been his main
interest. His strongest ability and strongest motivation is in
theoretical work. His experimental ability and motivation have
developed greatly. He has potential to do significant research after
his thesis. He is about to finish his PhD in computer science, and
is actively working toward completion. He appears on schedule to
complete his thesis in mid-summer. His thesis in artificial
intelligence concerns the interpretation of scenes of polyhedra from
arbitrary pairs of stereo views, particularly wide angle stereo
pairs. Previously, he had worked on monocular interpretation of
scenes of polyhedra from an orientation of heuristic body separation.
He has recently shown a strong interest in several areas of computer
vision, including biomedical imaging. He has also shown a recent
interest in the combinatorial problems of computer graphics and
collision avoidance for a manipulator moving in an environment with
obstacles.
His general knowledge of work in AI is adequate,but not broad. His
strongest interests have been in combinatorics. His knowledge of AI
vision work is adequate and a bit narrow; probably the same as
typical PhD holders from Stanford and MIT. His programming
experience and skill are adequate. His ability to make original
contributions to an AI project is strong. His interests have been in
quite formal and technical problems. His strength is in theoretical
analysis in combinatorics. He has worked largely alone and has
recently shown motivation and energy. He has little knowledge of
related fields. His ability to document his ideas and programs seems
adequate.
I have not seen him teach. He is easy to get along with, and has no
obstacles to free and easy communications with colleagues. I presume
that he could work in a team of four or five people. However, he has
worked alone and seems to prefer that.
Previously, you had requested a reference for Arthur Thomas. Between
the two, Ganapathy is currently working in vision. Both are good. I
consider Arthur Thomas to be far superior, in regard to his knowledge
of AI in general, in his knowledge of vision, in energy and
motivation, in his knowledge of related fields, in his ability to
document ideas and programs, and in potential for original
contribution in vision. The two are about equal in programming
experience and skill.
With best regards,
Thomas O. Binford
request for Turner thesis
Stanford Artificial Intelligence Laboratory
Computer Science Dept
Stanford University
Stanford, California,94305
Prof Rod Burstall
Dept of Machine Intelligence
University of Edinburgh
9 Hope Park Square
Edinburgh, Scotland
Dear Rod:
I am writing to ask a favor. I am to give the vision tutorial for
the IJCAI in USSR. I'd like to include something on Ken Turner's
thesis. Funny thing, I haven't gotten a copy yet. Harry Barrow has
told me of the hangups in publication. I'm writing to ask something
nonstandard. Could you get one of your assistants to xerox a copy
for me, and bill us for the expense? I'd appreciate that a lot.
I hope to see you this fall. I don't know whether you have heard
much about the funding situation with ARPA. We received small cuts
and only a six month contract. We will have to go through the
proposal trauma almost immediately. I have gotten a small additional
contract, and we are more than half on NSF (RANN) funding, so we are
not in bad shape. Dave Grossman from IBM will be here in August, to
work on the automation project. AL, the language for assembly will
be working in zeroth version about the end of June. As far as
funding from ARPA, they seem to want a complete reorientation to
military applications. As far as I can see now, the military
applications are hard to evaluate, and look difficult. What we can
do best is to basic things which will be useful for what they want,
but they'll be useful for other things, too. I'm convinced that
vision is really going to pay off on the ten year scale. Partly,
because I see a factor of 1000 in state of the art over PDP10 in
computing power for vision (tenfold ECL 8 bit multiply in 50nsec),
another factor of two per two years gives 1000 in twenty years and a
factor of 10 in gate speed. That's a factor of 10ā7 in twenty years.
Now we avoid a lot of things because they require computation, but
that's inherent in vision. The human visual cortex is at least six
orders of magnitude more powerful than the PDP10 for vision. We
probably will get fast before we get smart.
Sincerely,
Thomas O. Binford
Yasuda
Dear Mr. Yasuda:
Thank you for your letter. I enjoy your interest in scene analysis, and
will mention what we are doing and what references are available.
I will offer a course in the fall:
CS 227 Robotics
The course covers scene analysis, manipulation, languages for vision
and manipulation, and planning systems (very high level language)
The sort of work we are doing is connected with three dimensional scenes.
I would characterize it as "DESCRIPTION, NOT RECOGNITION". That is, the
essential effort is to describe visual elements and relations in 2d and
3d descriptive terms, rather than classify among a few models. Thus, we
have a different orientation from classical pattern recognition. There
are no adequate books. I will suggest some articles and theses:
Our research areas are as follows:
VISUAL SENSING FOR ASSEMBLY
We use CAD (computer aided design) models with graphics, or training images
to sense alignment errors for repetitive assembly situations. The objective
is a programming system for simple programming of a class of problems, not
a single problem.
Features, modelling, strategy, stereo, correlation, graphics
SCENE ANALYSIS FOR ASSEMBLY
Vision applied to more complicated problems in assembly and automation.
The above problems are 2 1/2 dim problems, quasi two-dimensional. The
typical tasks here are picking objects out of a bin.
VISUAL CONTROL OF NAVIGATION FOR PLANETARY EXPLORATION
Visual sensing of terrain and navigation using a computer controlled cart
with a television transmitter. Stereo terrain mapping by motion parallax.
REPRESENTATION OF SHAPE
Representation of shape of complex objects is used in planning systems for
vision and assembly, and is used in perception and graphics/man-machine interfaces.
We have an effective representation which we are implementing in several ways.
SHAPE DESCRIPTION OF COMPLEX OBJECTS
We have recognized objects of the complexity of dolls and toy horses, using
descriptions based on the above representations, and using a structured visual
memory.
VISION LANGUAGE
Representation of 2d and 3d structures and relations, representation of control
structures, representation of operations
CARTOGRAPHY, PHOTOINTERPRETATION
This is a new direction for us. Interactive aids to map-making, etc.
It is related to our past work in interactive image processing for MARS missions.
BIOMEDICAL IMAGING
We are just beginning to make a program in this area.
COMPLEX OBJECTS and REPRESENTATION
Nevatia and Binford 3rd Intern Joint Conf on Artif Intelligence
Agin and Binford 3rd Intern Joint Conf on Artif Intelligence
R.K.Nevatia, thesis, Comp Science Dept, Stanford Univ, 1974
Miyamoto and Binford, Computer Graphics and Image Processing Conf, May 1975
PERCEPTION AND LOW LEVEL VISION
David Marr, MIT AI Memos
David Waltz, MIT thesis, AI
Thomas and Binford, Stanford AI Memo
special issue of SIGART newsletter, July 1975
May I suggest that you also talk with Dr. Y. Shirai, at the Electrotechnical
Laboratory, Tokyo. Dr. Shirai is a distinguished researcher in vision.
Sincerely,
Thomas O. Binford
Thomas
Stanford Artificial Intelligence Laboratory
Computer Science Dept
Stanford University
Stanford, California,94305
Prof Aaron Sloman
Cognitive Studies Programme
School of Social Sciences
The University of Sussex
Brighton, BN1 9QN
England
Dear Prof Sloman:
I am writing in response to your request for a reference on the
qualifications of Arthur Thomas for the post in your vision project.
His knowledge of AI is very broad; he reads very widely and has an
unusual command of the literature. He has a detailed knowledge of
the work in vision and in mathematical theory of computation. He has
a fine knowledge of work in vision; he is perhaps the most
sophisticated of vision researchers at the doctoral level. He has
made substantial contributions in perception of color constancy and
in depth perception based on motion parallax of selected features.
As you know, his thesis work has been in mathematical theory of
computation, formal reasoning. However, I would rank him as the
outstanding young vision researcher currently available. As an
overall evaluation, he is outstanding. He is very bright, has a very
wide grasp of technical literature, has great energy and very strong
motivation to basic science issues. His orientation has been more to
science than very narrow applications. He has expressed strong
interest in applications. He has a good but not exceptional
experience and background in programming, and learns very quickly.
He is very adventuresome and masters a lot of programming. He has
great ability to make original contributions to an AI project. He
writes extensively, articles, manuals, and reviews. He has an
exceptional knowledge of related disciplines: psychology of
perception, philosophy, and of course, formal logic.
He would be a distinct advantage to a team. He is well-liked by
others here in the laboratory, and has worked successfully with
several different people in our group.
Sincerely,
Thomas O. Binford
Ganapathy
Stanford Artificial Intelligence Laboratory
Computer Science Dept
Stanford University
Stanford, California,94305
Prof David Moursund
Computer Science Dept
University of Maryland
Eugene, Oregon, 97403
Dear Prof Moursund:
I am writing to support Sundaram (Kicha) Ganapathy's application for
a position with you. He seems one of the brightest students we have
had in our group. He has specialized in computer vision and has some
related research interests as well. He has shown strong interests in
theoretical problems in combinatorics, and produced some results, in
addition to his work in computer vision. He has done considerable
programming, although that has not been his main interest. His
strongest ability and strongest motivation is in theoretical work.
His experimental ability and motivation have developed greatly. He
has potential to do significant research after his thesis. He is
about to finish his PhD in computer science, and is actively working
toward completion. He appears on schedule to complete his thesis in
mid-summer. His thesis in artificial intelligence concerns the
interpretation of scenes of polyhedra from arbitrary pairs of stereo
views, particularly wide angle stereo pairs. Previously, he had
worked on monocular interpretation of scenes of polyhedra from an
orientation of heuristic body separation. He has shown a strong
interest in several areas of computer vision, including biomedical
imaging. He has also shown a strong interest in the combinatorial
problems of computer graphics and collision avoidance for a
manipulator moving in an environment with obstacles. Compared to
other students in our group, I would estimate that he is among the
best in theoretical work and average in experimental work.
I have not seen him teach. He is easy to get along with, and has no
obstacles to free and easy communications with colleagues.
With best regards,
Thomas O. Binford
Ganapathy
Stanford Artificial Intelligence Laboratory
Computer Science Dept
Stanford University
Stanford, California,94305
Prof Azriel Rosenfeld
Computer Science Center
University of Maryland
College Park, Md, 20742
Dear Prof Rosenfeld:
I am writing to support Sundaram (Kicha) Ganapathy's application for
a position with you. He is about to finish his PhD in computer
science, and is actively working toward completion. He appears on
schedule to complete his thesis in mid-summer. His thesis in
artificial intelligence concerns the interpretation of scenes of
polyhedra from arbitrary pairs of stereo views, particularly wide
angle stereo pairs. Previously, he had worked on monocular
interpretation of scenes of polyhedra from an orientation of
heuristic body separation. We have talked about what areas of work
he would like after finishing his thesis, and he has shown a strong
interest in several areas of computer vision, including biomedical
imaging. He seems one of the brightest students we have had in our
group. He has shown strong interests in theoretical problems in
combinatorics, and produced some results, in addition to his work in
computer vision. He has done considerable programming, although that
has not been his main interest. His strongest ability and strongest
motivation is in theoretical work. His experimental ability and
motivation have developed greatly. He has potential to do
significant research after his thesis. You asked for comparative
ranking with other students in our group. I would rank Nevatia
significantly higher. I would rank Agin higher in experimental work
and Ganapathy higher in theoretical work. Overall, he would rank at
about 40% where 100% is our best.
I have not seen him teach. He is easy to get along with, and has no
obstacles to free and easy communications with colleagues.
With best regards,
Thomas O. Binford
Ganapathy
Stanford Artificial Intelligence Laboratory
Computer Science Dept
Stanford University
Stanford, California,94305
Employment Dept
IBM Thomas J. Watson Research Center
P O Box 218
Yorktown Heights, New York, 10598
Dear Sir or Madam:
I am writing to support Sundaram (Kicha) Ganapathy's application for
a position with you. He is about to finish his PhD in computer
science, and is actively working toward that and seems motivated.
His thesis in artificial intelligence concerns the interpretation of
scenes of polyhedra from arbitrary pairs of stereo views.
Previously, he had worked on monocular interpretation of scenes of
polyhedra. I have known him over a period of several years. He is
one of the brightest students we have had in our group. He has shown
strong interests in theoretical problems in combinatorics, and
produced some results, in addition to his work in computer vision.
He has done considerable programming, although that has not been his
main interest.
I have not seen him teach. He is easy to get along with, and has no
obstacles to free and easy communications with colleagues.
With best regards,
Thomas O. Binford
Miyamoto, April 16,1975
Stanford Artificial Intelligence Laboratory
Computer Science Dept
Stanford University
Stanford, California,94305
Dear Prof Miyamoto:
I am very glad that you will attend the conference. I prefer very
strongly that you deliver the talk on the paper. My strongest motivation
is that others have a chance to see you and come to know you.
I enclose a copy of the text of the paper. I don't have a copy with
the diagrams, because that was to be reduced by 3/4 and won't fit on a
Xerox sheet.
Mr. Takeda has sent a copy of the program of the National Computer Conference.
I look forward to seeing you. Please call me and let me know how to reach
you on your arrival.
With best regards,
Thomas O. Binford
John McCarthy April 16,1975
Dear John:
I want to repeat the information about a couple of hand/eye matters
and negotiate with you. There is some new information about Meir
Weinstein. He says that the students will be supported by Caltech,
and we could support a couple of students and some lab overhead.
1) Meir Weinstein contacted me to negotiate about moving here with
$100k. I'll give you the facts as I know them, and my opinion. He
has $40k from NASA and $60k from Bill Whitney. He thinks he can
bring both along. If he can't bring both, then financially we have
no interest. Figure $40k for Meir and $20k for two students,
according to Les, and that leaves $40k for overhead.
What do we get out of it? My opinion is that Weinstein would be
occupied with his own things and would not be of great utility. He
would probably continue to supervise three Caltech students remotely.
Fredkin thinks of him as not having produced anything original, but
he has right ideas. I agree and do not expect much.
He wants to change because Caltech has decided not to have robotics;
Fredkin confirms this. He is negotiating for some other
possibilities at Caltech, also Irvine and UCLA. He has a fallback at
JPL. His priorities are Caltech, Stanford, then Irvine or UCLA. He
wants to come July 1 for two years, after which he will go to Israel.
He will accept whatever appointment is easiest, preferably a research
appointment.
Are there any risks? He might want to stay and do some politicking
to stay. What happens when his money runs out? Would he be using
resources at cross purposes?
I recommend that we NOT do it under these conditions. We get
marginal value.
2) Masatsugu Kidode of Toshiba has written asking about possibility
for visiting here for a year and a half. He has worked for five
years on character recognition (at least one Toshiba system I saw had
a reasonable approach, as contrasted to most) and automatic cancer
cell detection. He wants to do research on satellite and aerial
imagery. He does have support. I will follow it up. If you like,
you might get in touch with him. I met him but did not have an
evaluation. I WANT HIM. There is always a question about his English.
PLEASE CALL ABOUT MEIR.
With best regards
Tom
Kidode April 7,1975
Stanford Artificial Intelligence Laboratory
Computer Science Dept
Stanford University
Stanford, California,94305
Dr. Masatsugu Kidode
Information Systems Laboratory
Toshiba Research and Development Center
1,Komukai Toshiba-Cho,Saiwai-Ku
Kawasaki-City
Kanagawa 210, Japan
Dear Dr. Kidode:
I strongly support your coming here as a visiting researcher, paid
by Toshiba. Research in satellite and aerial imagery is interesting
and timely for us. I have written Prof John McCarthy who must
approve. You might contact him c/o Prof Takase, Kyoto.
I am glad to hear from you again and hope that all is well with you.
With Best Regards,
Thomas O. Binford
John McCarthy April 7,1975
Dear John:
1) Meir Weinstein contacted me to negotiate about moving here with
$100k and three students. I'll give you the facts as I know them,
and my opinion. He has $40k from NASA and $60k from Bill Whitney.
He thinks he can bring both along. If he can't bring both, then
financially we have no interest. Figure $40k for Meir and $30k for
three students, according to Les, and that leaves only $30k for
overhead, which is less than we have on other contracts.
What do we get out of it? Three students: Srinivas-error recovery
and representation of plans; Udupa-collision avoidance; Ross-LISP
hacker and possibly vision. My opinion is that Weinstein would be
occupied with his own things and would not be of great utility.
Fredkin thinks of him as not having produced anything original, but
he has the right ideas. I agree and do not expect much. We also get
some overhead, but not much.
He wants to change because Caltech has decided not to have robotics;
Fredkin confirms this. He is negotiating for some other
possibilities at Caltech, also Irvine and UCLA. He has a fallback at
JPL. His priorities are Caltech, Stanford, then Irvine or UCLA. He
wants to come July 1 for two years, after which he will go to Israel.
He will accept whatever appointment is easiest, preferably a research
appointment.
Are there any risks? He might want to stay and do some politicking
for that.
I recommend that we don't do it under these conditions. We get
marginal value. Are there any conditions that are attractive?
2) Masatsugu Kidode of Toshiba has written asking about possibility
for visiting here for a year and a half. He has worked for five
years on character recognition (at least one Toshiba system I saw had
a reasonable approach, as contrasted to most) and automatic cancer
cell detection. He wants to do research on satellite and aerial
imagery. He does not say that he has support. I will follow it up.
If you like, you might get in touch with him. I met him but did not
have an evaluation. If he has support, I want him; otherwise, not.
3) An EE who is on leave from DuPont, Mike Roderick, is going to work
free for us on arm control. He is going to do z-plane analysis of
control, to see what sample rates are possible. The idea is to cut
the computation load. In my estimate, something good will come out
of it, even though computing hardware will probably be cheap soon,
and the motivation to cut computing is not so important as it seems.
For the next couple of years that might be important. Lou thinks
this is a good thing to work on. He thinks that Roderick is good.
Roderick works in the area for DuPont and the project would be good
for him. I have agreed to his working here, using the facilities.
4) In the light of the financial problems and the threat of worse
problems, my own plans for the hand/eye group are to cut support
including hardware as much as possible, to find how we can fit into
ARPA programs, and to seek other support. for vision for vehicles
and aerial imaging. In particular, I am now evaluating biomedical
and biological imaging. What I want is an alliance with a biological
group; I don't think that out of our heads we can conceive advances
in biology. I do want to know enough to evaluate whether something
seems like a good idea to me. I don't want to get involved in any
bad ideas. I think there are enough good projects around.
Please call about Meir
With best regards
Tom