5 / 24
meeting now which is different from
before?
So there is a change in the number of
people because now we have many
more attendees that are not
presenting. This used to be a
conference where everybody was
presenting something. What has
changed now is that we have a number
of companies that are coming that
have a booth, and they are presenting.
Their role here is not to come and
present research, but rather to attract
talent. That has changed. What I’ve
seen also is a difference in the way we
present results. Today we have large
datasets where you can measure
progress. I’m kind of a pioneer from
back in the old days of computer vision
when you created your own datasets.
You presented your results, and people
had to evaluate the quality and the
value of it in a very non-scientific way,
in a sense. Today, what you have are
datasets. You can measure 50% this or
10% that. That is a measurable
difference which allows us to say,
“
Algorithm A versus algorithm B: I
know how these two can compare, and
this is what we should be going with.
”
That has allowed the field to move
much faster than it moved before.
Of all the things that you’ve heard in
the last few days, is there one “wow”
moment that struck you in particular?
Ah, that’s a difficult question. I like,
very much, the oral papers. I’m not
going to cite them. I think the award
papers are very much deserving. Every
one of the award papers was different
and brought a new perspective, a new
idea that was really not known before.
That’s why they got the award. I think
the award committee did an excellent
job of selecting the papers. That said, it
is sometimes difficult to evaluate the
impact of a paper that is being
published. There are awards now for
papers that have stood the test of
time. That, to me, is valuable, to look
back, not just at this conference, but
let’s go back ten years. Ten years ago,
which paper actually made a difference
that people are citing today and are
using today?
What impresses you when you see a
presentation or you read a paper?
There are different ways that I can be
surprised. One is a result that I didn’t
think we could achieve. There was a
paper from MIT that talked about
looking around the corner.
Of course, by
Katie Bouman !That’s right. That’s a surprise. That’s
something where you say, “
I didn’t
think that you could do something with
these types of images
.” Again, they
show that, yes, there is information
that we can extract. That’s one type of
surprise. The other type of surprise is
producing a rich caption from an image
or from a video. Five or seven years ago,
5
Friday
Gérard Medioni