COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	∂14-APR-75  1821		network site ECL
C00013 00003	∂22-FEB-75  0404		S,LES
C00021 ENDMK
C⊗;
∂14-APR-75  1821		network site ECL
 Date: 14 APR 1975 1821-PDT
 From: NEVATIA at USC-ECL
 To:   TOB at SU-AI
 
 Hi Tom, Thanks for your note.
 Re linking of edges: the info I was looking for was the linking of
 edge elements for blocks scenes, such as used in Gunnar's thesis
 (he does not describe his algorithm). I presume similar techniques
 have been used in other blocks programs (not including line following).
 None of your references was to a blocks program!
 Further, I was interested in connecting  long, linked edges into
 object boundaries. For blocks scene, this is easily accomplished
 by fitting straight lines and finding intersections etc. I was
 looking for a more intelligent way of connecting the segments,
 based on some notions of outside of an object and restrictions
 on the smoothness.   I know of no such attempts, do you?
 
 On color edges, I have been in touch with AJT. I was more interested
 in color "edges" as you defined it, and was looking for a program
 to see how you fitted the "moments" of three components. I have
 some trouble finding the moments in the output of Heuckel, and
 was looking for a short cut by looking at some existing program.
 
 I have some good results on linking Heuckel output in straight line
 segments for the tank scene. Connecting these segments into a complete
 boundary, however, seems to be tricky.
 
 Going to Maryland conf. tomorrow. Talk to you when I return.
 
 Ram.
 -------


∂12-MAR-75  1800		network site ECL
 Date: 12 MAR 1975 1800-PDT
 From: NEVATIA at USC-ECL
 To:   TOB at SU-AI
 
 1. I will be up at AI lab, this Fri (Mar 14). Would like to
   see you for some time. Have some thoughts to discuss  and also some details to be cleared.
 2. Tried to load your lateral inhibition edge detector (local.5t),
   got some undefined globals using vision.11 and undeclared variables
   using a later version of vision. It is perhaps just a small bug.
   I was hoping to see its results on some images.
 Hope to see you on fri.
 Ram.
 -------

∂10-MAR-75  1848		network site ECL
 Date: 10 MAR 1975 1830-PDT
 From: NEVATIA at USC-ECL
 To:   TOB at SU-AI
 
 Hi Tom,
 	I could use some of the Hand-Eye routines right away. Mainly,
 I like to use the picture file input and output routines. In the mean
 time I have written some of my own, but they are just quick hacks.
 I may come up to SU later this week. If someone can look into it
 before, it would make my job easier, else I will struggle with it then.
 Our paper got accepted to the Conf. at Maryland. No written paper is
 required. I think that I will go. Were  you planning on it too?
 Ram.
 -------
 

∂4-MAR-75  1357		network site ECL
 Date:  4 MAR 1975 1401-PDT
 From: NEVATIA at USC-ECL
 To:   Binford at SU-AI
 
 Tom, I want to use some of the Hand Eye library routines and need
 their source files to compile here. I understand that Karl Pingle
 has left. Is someone else taking over his function and knows about
 such things?
 Your new Vision Language write up is much better. I still have trouble
 with data structures and details of what exactly does each operation do.
 Please send me a note if someone can help with the Hand Eye stuff.
 (If you have problems with Network mail, leave a note on the 
 Stanford system.)
 Thanks.
 Ram.
 -------

∂22-FEB-75  1510		THE,AJT
 you mentioned that you were cataloguing yr. library. BIB.SAI[BIB,AJT]
 is a program which prompts you for information, and makes up the
 TFL file that is used by INDEX to  reformat and keyword files for
 the NS. MAKBIB shows you how to run index, READBIB allows you to
 use the NS (these are both DO files on the same area)
 To use BIB.SAI, copy it over, change the filenames &c, and load it.
 When run,  it puts stuff into a TFL file if one exists (formatted, with
 `∂' as story delimiter) or makes a TFL file. Theeprompts it gives
 should be self explanatory. If you want multi-line input to either the
 TITLE or SOURCE, end all your lines except the last with `~'.
 To exit, say <cr> to the Book,reprint,citation question.  This program
 saves the TFL file every five entries (untrusting, isn't it) and then
 continues.
 enjoy. a.
 ps. remember - if you use this, you must NEVER edit the .TXT file that
 INDEX generates - thats one reason why MAKBIB makes a separate copy called
 .DOC
 pps. I forgot - you also need a file called INUM.DAT in which the
 sequence number is stored (start it out by making INUM.DAT just 
 containing a 1 (or 0, I can't remember and it doesn't matter)

∂17-APR-75  1353		S,LES
 Regarding the planned visit by Dave Russell of ARPA next Wednesday,
 I propose the following schedule.  Please let me know if you would
 like a different time slot.  I guess that we can do most of it in
 my office.
   9-9:30 am	LES
   9:30-10	TW
   10-10:30	TOB
   10:30-11	RWW
   11-11:30	CCG
   11:30-11:50	Demos & bullshit
   12-1 pm	Lunch @ Faculty Club
   1-2		Feigenbaum demos
 
 Let me know if you would like to go to lunch.  We might even pay for it.
 Tom:  it might be good to give a hand-eye demo.


COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	∂22-FEB-75  0404		S,LES
C00003 ENDMK
C⊗;
∂22-FEB-75  0404		S,LES
 The ARPA proposal segments are located in <proj>.PUB[R,LES],
 where <proj>= AI, FR, AD, PU, NL, & HE.  You may snarf your
 part or edit in place.
 
 The first line of each of these files is
 .REQUIRE "HEAD.PUB[R,LES]" SOURCE_FILE;
 which sets up PUB macros for two-column output.  Thus, you can
 Pub your part to see what it looks like.  For a brief description
 of what the principal macros do, see Page 2 of HEAD.PUB [R,LES].
 
 The corresponding segments of our last annual report are located
 in files <proj>.RPT [R,LES].  Have fun!
 					Les
∂05-APR-75  0756		1,KIC
 Tom, There are 2 things i wanted to cover with you.Both pertain to my future
 plans - short and long.First the long
 (1)	I mailed off my resume to Rosenfeld and spoke to Sklansky at Irvine on
 the phone .He seemed very interested and wanted me to come down for an interview
 as soon as i could.I postponed it to early May (soon after the Orals).He wanted to
 know  a lot.I expect you  might get a call from him.I have sent him also my resume.
 I am applying to a few other places and would appreciate very much any help you
 can render.
 (2) 	This brings me to my short range plans.Needless to say ,i have been working 
 very hard for quite some time now and am back in the groove or state of mind with
 which i acme to Stanford 5 Yrs ago.In view of that i was optimistic of full time
 support over summer and thought i would get done TOTALLY by end of Aug and take off
 to India for a mont or so.But after talking to you and hearing Les at the Lab
 meeting i am convinced  that the sooner i take off from here the better off i
 would be.Now i have stepped up my pace even more but it looks difficult that i
 would be able to submit my thesis be end of June.End of July or mid Aug seems
 like a viable date ,considering the fact that the readers have to give me their
 comments and i have to make corrections etc.I do not like the idea of leaving Stanford
 and submitting my thesis later.I,on my part though expect to have the thesis
 completely documented by June.
 	I gave you 10 days ago 3-4 page writeup on real scenes and how to improve
 things.I am perfectly willing to work on anything related to my Thesis to earn my
 support over summer.I do not wish to be a dead weight nor am i expecting to be
 supported just for writing a thesis.In view of that i would very much appreciate
 half time support over Summer on Automation project or in fact anything else.You
 know me well enough  to know what things i am capable of doing and what not! 
 	It seems ironic and unbeleivable that i cannot be supported for 2-3 months
 over summer at my current pace.The reason i want to stay on at the lab till i finish
 is because of Visa restrictions.I cannot just go and take up a part time job somewhere
 else outside the university.Fulltime is something else if i am willing to change my
 Visa status.In view of all these i hope you can find resources to support me over
 summer.If you or Les needs to talk to me about this i would welcome it.I am sending
 you this note instead of talking to you because i thought it would be a good idea
 to make my request in writing. Thanking you in advance(!)		Kicha
 P.S   	Please confirm receipt of this note. On your final decision,i can wait for
 a couple of weeks before i start looking elsewhere.

∂13-APR-75  0727		network site CMU
 **** FTP mail from [X180LE03] (SIGART)
 o:	Tom Binford (TOB @ SU-AI)
 rom:	Lee Erman (SIGART @ CMU-10B)
 
 om,
 
  have two requests:  First, I would appreciate if you could send
 e a copy of your paper "Information Processing Analysis of
 isual Perception".
 
 econd, as editor of the SIGART Newsletter, I would like to ask
 ou to participate in next issue's special topic, dealing with
 achine visual perception.  If you are not familiar with this
 atest gimmick, I'm appending some stuff from the last Newsletter
 elow.  We are trying to encourage as many of the "good" workers
 n the area to participate -- anything you could submit (or con
 thers into submitting) would be very much appreciated.
 
 hanx,
 Lee
 
 nclosure:
   
   II.1.  Technical Topics
       As  described in  last issue's  Editors' Entry,  each  issue will
   feature a technical topic.    For June, the topic is  Computer Vision
   and Perception and, in particular, addresses the following questions:
   .What knowledge is necessary to segment and interpret scenes?
   .How should such knowledge be represented and used?
   .Are there methods general enough for a wide class of scene types?
   Jay  M. Tenenbaum  and Harry  Barrow, of  SRI, will  write  a feature
   article  on  this  topic; others  are  strongly  urged  to contribute
   answers to these questions.   All submissions received by May  1 will
   be  circulated to  all those who  have contributed  by then.  "Final"
   submissions for the Newsletter are due May 20.
 ---