To facilitate the display and discussion of image sequences, CVPR '97
submissions may refer to Web-served sequences.  This is intended for
sequences only; for this year, static images should still be included
in the document directly.  This is an experiment whose efficacy will
be discussed at the meeting.  

The guidlines for use are as follows:

1) You may NOT assume that the reviewers will look at the sequences.
Yes, we know that is annoying, but we are still unwilling to require that
reviews be on-line while reading papers.  Thus the Web-based sequences
should be for additional clarity; the paper should stand without them.

2) In order to maintain double-blind reviews, the submissions should
refer to "http:[www-seq-]" for each seqeunce x (please let x start
at 1).  For review purposes, we will build a site that is indexed by
the assigned manuscript number, that contains a *copy* of the
sequences.  In order for the this to be set up, you must send an email
to BOBICK@MEDIA.MIT.EDU giving the URL of each sequence to be copied.
Make sure the order in which they listed correspond to the order
referred to in the manuscript.  The seqeunces will be downloaded to
create the review site.

*** This COPY site will not be protected from the public, though we
will try to make them crawler resistant.  The files will only have
manuscript number to indicate origin.***

3) Space limitations: Seqeunces should be QT or MPEG (MPEG is probably
more universal for workstations, QT for PC/Macs).  For now, we are
setting a limit of 30MB per submission.  If this limitation is a
problem send a note to BOBICK@MEDIA.MIT.EDU.

4) If your paper is accepted, it will be you obligation to change the
URL listed in the paper to be your real site.  The review site will
dissappear shortly after the acceptances are announced.

Questions: send email to Aaron Bobick, bobick@media.mit.edu