CVPR Daily - Tuesday

6 / 30

I started my PhD 17 years ago, and one
way to keep the passion over 17 years
is to change the family of problems
every once in a while, to freshen up.
Stay in machine learning and computer
vision problems, but change at the
beginning every 3 years and later every
6 years, and eat from the diversity of
the computer vision fruit - it’s a really
big fruit.
Is that something that others do as
well?
If I may dare to mention the really
great, like Zisserman and Malik, who
manage to keep the passion for a long
time, they all have one distinguishing
mark: they worked on a lot of
problems, and typically they make one
landmark contribution in each era.
Can it happen that somebody enters
into a field and realises that they
shouldn’t have?
Oh, absolutely. It happens when you’re
younger, and it happens when you’re
older. You have to be able to feel
whether putting energy into an area is
going to lead to things you want.
Which are always the same two:
happiness for yourself, so you have
fun, and the second is your publishing
and that people are interested in what
you write. These two criteria are often
in contradiction. So you need to feel it
as fast as you can. I would say if you’re
not happy after six months after
entering a field, you should change.
How do you rebound in the case it
doesn’t work?
When you are the first implementer - a
PhD or a postdoc, before you are a
professor, rebound is somewhat easier.
You have to have the self-discipline of
going to your advisor and saying, look,
this thing doesn't work. And then
normally it’s about having a picture of
the new thing. So just saying “I hate
what I’m doing, and I quit because it
doesn’t work”, then the alternative is
the void, and the void scares
everybody. So you need to set aside
some time. Normally when I felt I
wasn’t doing so well in an area,
especially when I was younger, I would
just say: this week, I only read. I don’t
program anything. And just read as
diverse stuff as possible, and then
decide what to work on. When you are
a older and you are a group leader,
then it’s harder to rebound, because
you are very much in love with your
own vision [he laughs] and you don’t
quite see why it doesn’t work.
And you have a responsibility for the
people who are with you.
Yes, telling your students: you know
what, because of various reasons,
perhaps technical impossibility reasons
or because somebody else already
implemented your idea, you have to
change direction. This is tricky, but it’s
important to do it. As you said, you
have a responsibility for the student.
And sometimes the best interest of the
student is to radically change topic. As
a group leader you must make these
choices.
Might it also be the case where the
student opens the eyes of the
professor and says that something is
not going to work?
Absolutely, sometimes it’s exactly the
PhD student or postdoc that has to go
to his boss and say: you know, Vitto,
this thing you like so much - it ain’t
gonna fly [he laughs ]. The professor
6 Vittorio Ferrari
“humble down sometimes”