Vision-List-Request@ADS.COM (Vision-List moderator Phil Kahn) (06/12/91)
VISION-LIST Digest Tue Jun 11 12:36:59 PDT 91 Volume 10 : Issue 27
- Send submissions to Vision-List@ADS.COM
- Vision List Digest available via COMP.AI.VISION newsgroup
- If you don't have access to COMP.AI.VISION, request list
membership to Vision-List-Request@ADS.COM
- Access Vision List Archives via anonymous ftp to ADS.COM
Today's Topics:
Associative memory for image retrieval
Cheap IJCAI91 airfares available (but not for long)
Position Wanted
BMVA Technical Meeting
DATACUBE info. responses (very long)
----------------------------------------------------------------------
Date: 7 Jun 91 10:51:55 GMT
From: wu@canon.co.uk
Organization: Canon Research Europe, Guildford, UK
Subject: associative memory for image retrieval
I am trying to collect information about the use of associative
memory for fast image storage and retrieval. Any reference to the
topic is welcome. I will summarize responses.
Many thanks.
S.F. Wu wu@canon.co.uk
------------------------------
Date: Thu, 6 Jun 91 18:00:42 EDT
From: mostow@cs.rutgers.edu
Subject: Cheap IJCAI91 airfares available (but not for long)
For those of you planning to attend IJCAI-91 in Sydney, Australia, it
may be important to know that there are extremely discounted
round-trip airfares to/from the US and Sydney available right now.
Both Continental and Northwest are offering $685 (+ tax = $703)
round-trip fares from JFK and Newark. Both these fares *expire soon*
(Continental on June 7th and Northwest the following week) so, if
you're interested in taking advantage of the savings, call your travel
agent immediately!
Your local travel agent should have further details and be able to
make the arrangements for you.
I DON'T HAVE ANY OTHER INFORMATION, SO PLEASE DO ***NOT*** CONTACT ME.
------------------------------
Date: Wed, 5 Jun 91 12:35:46 BST
From: UCL Image Processing Group <ucipg@nsfnet-relay.ac.uk>
Subject: Position Wanted
POSITION WANTED
I am seeking a software engineering position in the field of computer vision
with a U.S. based company or research institution. I am a U.S. citizen and
am about to complete a PhD in image processing. I have extensive experience
in parallel processing (both SIMD and MIMD) as applied to image processing
and computer graphics. Please contact me for further details.
PhD Image Processing University College London
MSc Computer Science Rensselaer Polytechnic Institute
BSc Computer Information Science University of Massachusetts
Shawn Javid
100014.1731@compuserve.com
------------------------------
Date: Fri, 7 Jun 91 12:21:17 BST
From: J.Illingworth@ee.surrey.ac.uk
Subject: BMVA Technical Meeting
BRITISH MACHINE VISION ASSOCIATION
and SOCIETY FOR PATTERN RECOGNITION
ADVANCED TECHNIQUES FOR IMAGE CODING
Technical Meeting to be held on Tuesday 16 July 1991
at British Institute of Radiology,
36 Portland Place, London.
Chairman: Professor Peter Saraga (Philips Research, Redhill.)
10.15 Arrive and coffee
10.45 Welcome and introduction.
P. Sarage, Philips Research.
10:50 Standards for digital video coding.
Dr O.J. Morris, Philips Research Labs.
11:25 Combinatorial Optimisation Techniques for Image Coding.
Prof A. Constantinides, Imperial College.
12:00 Image Coding and Low Level Vision
Dr R. Wilson, Univ of Warwick.
12.35 LUNCH
14.15 Block Based Fractal Image Coding.
Dr J. Waite, British Telecom.
14.50 Model Based Image Coding.
Prof. D. Pearson, Univ of Essex.
15:25 TEA
15:55 Model Based Coding for Low Bitrate Video Telephony.
Dr Stanger, G.E.C. Hirst Research.
16:30 Development of Motion Estimation and Compression
Schemes for Image Sequence Coding.
Prof. R. Clarke, Heriot-Watt Univ.
17:05 Close
Dr. J. Illingworth, | Phone: (0483) 300-800 Ext. 2299
V.S.S.P. Group, | Fax : (0483) 300-803
Dept of Electronic and Electrical Eng, | Email: J.Illingworth@ee.surrey.ac.uk
University of Surrey, |
Guildford, |
Surrey GU2 5XH |
United Kingdom |
------------------------------
Date: Tue, 11 Jun 91 10:36:23 EDT
From: nar@cs.wayne.edu
Subject: Summary of DATACUBE info. responses
Status: RO
Hi Folks,
The following is the summary of responses that I received about
DATACUBE systems. I thank the people who responded once again.
[ Lengthy responses were posted, and I am posting them below. However,
David Coombs noted that there is a List devoted to the Datacube
that is operated by Matthew Turk. In the future, please direct
all Datacube inquiries to the datacube-list:
"Everyone responded positively to the Datacube list idea,
so here it is! I think things are set up properly so that
mailing to "datacube-list@media-lab.media.mit.edu" will send
your message to the groups at MIT (the Media Lab and the MIT
AI Lab), Harvard, and Rochester. Hitting "r" (or the equivalent
in whatever mailer you use) should reply to everyone.
"I think we should feel free to have public Q & A's -- the
point is that I can benefit from your discussions (and hopefully
vice-versa). I will feel free to exhibit my ignorance of
Suns, Datacube, and life in general, if you folks will be
considerate and gentle in your responses! As always, we
should try not to burden each other with questions that are
answered in the first page of the manuals -- but also let's
keep in mind how bad the Datacube manuals are and realize
that sometimes we'll have missed something along the way.
"I had also promised a summary of MAGICAL. MAGICAL is a package
developed by Boeing (High Tech Center), which allows a Sun or Symbolics
user to control Datacube hardware, including a bunch of
commands to do initialization, I/O, image operations,
graphics, feature extraction, etc. If our lawyers ever get
their act together, we will be getting a preliminary copy
of MAGICAL to evaluate. At which time I'll offer my opinions.
It looks very promising, but I have many unanswered questions
about it. If you'd like a copy of the pre-release manual,
I'll send one, or talk to Charlotte Lin at Boeing.
"So that's all, folks. The ball's in your respective courts now!
Matthew Turk"
phil...]
**************************
Subject: DATACUBE
Here is some information about our DATACUBE setup.
We use two different setups.
setup 1)It is built around the following datacube boards (cost: approx 10K$can):
-Digimax: image A/D D/A digitizing board (frame digitizer)
-ROI-Store (2M): Region of interest frame buffer two-megabytes.
(used to store digital data acquired with Digimax)
-Maxware software.
This setup is used for image acquisition. The two VME-boards are mounted in
a 9-slot VME backplane. The backplane is connected to a SUN-3/110 through
a VME-to-VME bus repeater (PT-VME 910 by Performance technologies).
This setup works fine and is very reliable. I must point out that the Maxware
software will no longer be supported by Datacube and will be replaced by the
Imageflow software.
Datacube hardware works fine but requires some installation efforts!!!
Setup 2: It is built around the following boards (total cost: 45K$ can)
-Digimax: image A/D D/A digitizing board (frame digitizer)
-ROI-Store (2M): Region of interest frame buffer two-megabytes.
(used to store digital data acquired with Digimax)
-Maxware software.
-VFIR MK-II real-time convolver
-MaxMux for data switching and real-time table look-up
-Euclid general purpose signal processing DSP board.
-VME-VME bus repeater and Sun-3/110
We are currently porting this setup on a Sun-4 IPC workstation using a
S-bus to VME-bus repeater (Performance Technologies). It is not completely
re-installed however.
We are also contemplating the purchase of a Datacube University package B
(43% off) sold for 20K$US:
-7-slot VME chassis with power supply and nice carrying case
-Digimax
-Roi-store 2M
-2 out of the three following boards:
-VFIR MK-II
-Featuremax
-Max-SP (low cost signal processor)
-a 68030 motorola master board processor
-Imageflow software (with 5-day training course)
-OS-9 operating system
-Ethernet link
-SCSI hard disk drive
We are going to use this setup for a robotics application.
Hope this information will be of any help.
Sincerely yours
Dr Denis Laurendeau
professor
Computer Vision and Digital Systems Laboratory
Department of Electrical Engineering
Laval University
Quebec City, Province of Quebec
Canada, G1K 7P4
phone: (418) 656-2979, Fax: (418) 656-3594
e-mail: laurend@gel.ulaval.ca
P.S. 1 : I do not work for Datacube!!!
P.S. 2 : Laval University is a French Speaking campus attended by 30K students.
This explains the poor quality of my English!!!
**************************
Subject: VISION-LIST digest 10.25
We have been using Datacube MAXVIDEO hardware since the mid 80's and I still
feel that the MAXVIDEO line has one of the best cost/performance tradeoffs
available FOR HIGH PERFORMANCE systems, i.e. where deterministic image
processing at TV frame rates or faster is required. If you only need to do
morphological, convolutions, fourier transforms, etc. in the seconds or tens
of second time-frame, you might consider a hardware system with better
software support (see below) or a software only system. (Khoros for example.)
While the MAXVIDEO hardware is a advantage in some cases, the major difficulty
is the USE of that hardware, or equivalently the software. IMAGEFLOW is
Datacube's software solution to the control of their hardware. While it is
better than the preceeding MAXWARE product, there still seems to be a lot
lacking in terms of technical user support and support for real-time
(inter-frame or millisecond rate) operations.
I do not like OS-9 as an operating system despite (or perhaps because of)
years of use. OS9 is a peculiar mixture of DOS, UNIX and VMS commands which
is recommended only because there are no good alternatives for fast,
stand-alone operating systems on generic VME equipment. I much prefer
developing/debugging in a workstation environment with interactive control of
the MAXVIDEO hardware via a Bit 3 bus repeater. This gives me a good
debugger, a UNIX front end, and the ability to use familiar tools, utilities
and flags. If you are interested in system installation away from a SUN or
other UNIX workstation, I have heard good things about VRTX.
Back to using MAXVIDEO with a Sparc 2, I know that Bit 3 has a Sbus to VME bus
convertor and while I haven't used it, the convertor/repeater approach would,
in my estimation, be preferable to learning and having to support OS9. Check
with Bit 3 (612) 881-6955 and SUN to see if SPARC 2's SUNOS comes with VME bus
drivers.
My recommendation is to use a SUN or whatever workstation with the MAXVIDEO
hardware tied "directly" (via Bit 3) to the workstation backplane. To
effectively use MAXVIDEO software, interactive control of the software is also
essential, which is NOT directly supported by IMAGEFLOW. At least one third
party vendor (below) provides easier to use environments than IMAGEFLOW. We
rolled our own enviornment orginally which I do not recommend if at all
possible.
Adaptive Automation - makes REAL - software access library for MAXVIDEO boards
(203) 289-9356
**************************
Subject: Info. needed about DATACUBE
I'm a programmer/analyst for a group at Cedar-Sinai hospital which is
using DATACUBE products to build a workstation environment to study
perception. We also use the boards to digitize RS-170 signals. We
have a digimax (RS-170 D<->A), a maxview D to A (combines sun-video
with the image stored in the viewmem), and a 2-Mb viewmem board.
We have the boards in a sun-4/370 (pedestal), though it would have
been worth it to get the card cage, since it is otherwise quite
difficult to remove boards ---all the boards have to be removed at
once since they're cabled together with a proprietary bus. And we had
to remove the boards multiple times, though once set up correctly
this is not a consideration.
The hardware manuals are buggy and difficult to use. Our field-help is
so-so. The software manuals are ok. It took us about a month and a
half (not full time) to get the hardware set up correctly and figure
out the right software parameters. System integration is not fun.
You might have to talk to Sun to find out certain details about the
Sun's video output timing, information which the boards must be given
via software. On the good side, though, the boards seem quite flexible
in terms of how they can be configured in hardware (busses) and software,
and the video signals they can deal with.
We have just begun using the boards in the last week or two, and do
not do image processing on them --we have no processing boards
right now, just the abovementioned digitizing and mixing boards.
Hope this helps.
David A. Honig
Email: honig@ics.uci.edu
c/o Dept. of Info. & Comp. Sci.
Univ. of California
Irvine, Ca. 92717
(714) 856-7310
c/o Dep't Med Physics and Imaging,
Cedar Sinai Med Center
8700 Beverly Blvd, L.A., Ca
(213) 855-2343
**************************
Subject: DATACUBE Hardware
We are running DataCube hardware using a Sparc 1+. The hardware is in
what they call a MAXBOX. It is connected to the Sun with a VME to
Sbus converter. We went the 'easy' but expensive way for using the
equipment. We went through a 3rd party systems integrator who has
developed their own code for accessing the stuff using UNIX commands
and C-callable libraries. Contact Steve Case at Inovision Inc. in
Research Triangle, NC for more information. His phone number is
(919)361-4609.
Brent Richburg
University of Illinois
Agricultural Engineering Department
1304 West Pennsylvania Avenue
Urbana, IL 61801
(217) 333-9415 office
244-4179 lab
244-0323 fax
**************************
Subject: Re: Info. needed about DATACUBE
I was a student at the University of Rochester. There we had many
datacube boards hooked up to a Sun 3/260 (with VME bus), and later a
Sun 4 also with a VME bus. Despite the work involved with dealing
with the incomplete and sometimes inaccurate documentation, the
Datacube was a great success, and a big part of our push towards
Active Vision. I personally used it to run my color histogram
algorithms. I believe that Greg Shiu at Wayne State has my thesis,
Color Indexing, which shows the Datacube setup for that.
Here at the University of Chicago, we have a Digicolor and a
Framestore, and will be getting a loaded Maxvideo-20 board, when
Datacube is able to make them right. (They've been telling us for
about two months that it's coming 'next week', but their production
problems have persisted.) The M-20 board combines a number of the old
boards into one double-width board at a significantly lower price.
(But note they've recently lowered the price on their old boards.) It
also has a cross-bar switch which means you can change algorithms
instantly from software instead of rewiring the cables ... which was
a real pain with the old boards. We're using a Performance
Technologies bus converter to attach the Bit3 VME cage we have to a
Sparcstation. That's supported by Datacube for their ImageFlow software.
We're just using Unix so far. I assume you have Glen
Ahearn's number at Datacube?
(508) 535-6644
Mike Swain
**************************
Subject: Re: Info. needed about DATACUBE
We have a bunch of datacube boards in a couple of VME chassis hosted
by a sun 4/280. (The sparcstation does not offer a VME bus.)
You would probably be interested in the new maxvideo-20 product, which
essentially combines the functionality of several of the older boards
on a double-wide VME board (mother board plus piggy-back). We have no
experience with the maxvideo, however.
Datacube's old software interface (MaxWare) to the maxvideo equipment was
awkward and buggy. I'm told the new ImageFlow software package is
much better, but being new still suffers from some bugs. We are using
some homegrown code, but will probably move to ImageFlow if we get
maxvideo-20s. ImageFlow is quite a piece of work, since it calculates
all the timing for the image paths in the system, so if it works it's
rather impressive. I just don't have experience.
Please let me know what you find and what you end up doing.
Dave Coombs
(716) 275-9098 or 5671
------------------------------
End of VISION-LIST digest 10.27
************************