Homepage of Arnold Smeulders
Published January 1st, 2010
Let me introduce myself. The driving factor has always been research up to the current day.
I say a few words about it below.
Vision has always interested me. How is it possible to communicate in words
about the content of an image? How is it possible to mutually decide what the
topic of an image is, when there is so much to see? How can the words be
learned to attach to the contents of an image? When the pictorial content of an
image is varying so widely from scene to scene and from instance to instance,
how can we recognize an object in an instance? Why is the one image considered
to be more beautiful than the other? How can we memorize pictures to the
thousands when so many bits are needed to capture them? How can pictures
express non-visual notions as happiness, democracy and tension?
So I am happy to be a scientist in computer vision.
Professor is the best job
in the world. First I spend a substantial part of my career in the medical
field of various universities in the Netherlands but later moved to the
computer science institute at the University of Amsterdam. Currently, we make
image search engines. Not by tracing what others think about the image - as
Google does very efficiently - but rather looking at the content of images. And
the computer vision community has recently learned how to successfully find
cars, planes and boats in an image. This may seem a trivial accomplishment but
remember babies spend their whole first year acquiring just this ability (and
not much more). In reality it takes half of the brain to process incoming
images so trivial it likely is not. With these search engines we perform very
well in the competitions for scientists posing harder and harder visual search
tasks every year.
Computer vision works much differently
than the human explanation of an object
they have seen. Then, they typically describe semantically important components
or functions of the object. But important components - such as the eyes in a
face - may be only a small part of the image. Or they may be difficult to
discriminate against other objects - there are many door-like things in the
world yet they are always used to describe a house. And, the function - a
chair is something to sit on - is not visual after all. It took a long time to
find out that vision should work by properties common to the type of objects:
boats are best described as a hole in the water because water shares many
visual characteristics and there are only a few other things which create a
hole in the water. When looking for a table knife, search for a table fork as
it has a very distinct shape common to all forks, then search with great
likelihood for the knife next to it. I suspect human vision works similarly,
surely before we start to talk about it. When we have recognized the object, we
construe intellectual decompositions by function and components.
The other thing is policy for the purpose of letting things blossom.
As chairman of the
ICT Platform Netherlands (IPN)
, we advise
the Netherlands organization for scientific research NWO
on the policy of ICT - science. The members of the IPN are representing important
research or educational institutes or groups of institutes in the Netherlands in
the full breadth of ICT. The mission of IPN is to represent and improve the condition
of ICT - science in the Netherlands. ICT is important; it is important in the economy,
in social life and in science. It is important by itself as it tends to formalize all
knowledge and information, and to decouple it from physical, temporal and location
presence. ICT drives change and innovation, in the end everywhere. As a general
purpose technology it is as important as the invention of electricity, raid roads,
machines and DNA. Therefore, we strive towards a solid representation of ICT - science
in the Netherlands by its own strength and in combination with practically all other
disciplines.
Another branch of policy making is
IIP/CREATE
.
The creative industry and
the creative lifestyle is important to the Netherlands. The service industry will
transform to the creative industry. Supported by ICT where all data will come loose
from their physical representation one can work and create and deliver services from
everywhere and every time. Large companies will be embedded in temporary networks of
individually operated endeavors. At IIP/CREATE we are trying to enhance the
traditionally poorly organized but famous Dutch creative industries with an emphasis
on ICT-driven components thereof (i-services and i- advertisements, digital media,
design and interface design, and alike).