Christophe Bourez
- 27 years old
- Researcher at INRIA
- Founder of Milpix
 Why did you go into this start-up venture?
Christophe Bourez: The start-up for image compression created
by Stéphane Mallat made a strong impression on me when I was
in the 2nd year of my Master's course in Mathematics, visualization
and apprenticeship at the Polytechnique. My stay in the United States
for my final course practicum at Princeton's Siemens research center
in the computer visualization department subsequently gave me an
opportunity to work around start-ups. This approach inspired me to
set up a project from A to Z. With the INRIA project-team LEAR,
I found the support I needed to see this approach through to completion.
These two factors played a major part in the birth of Milpix.
How did that happen? How can one say that the technology you are offering meets current needs?
Christophe Bourez: I discovered the work of Cordelia Schmid
at the Ecole Polytechnique. She was doing her research as part of
the INRIA project-team LEAR, a team that had considerable renown.
As of 1997, her research publications on local description of images
constitute a point of reference in the field of computer visualization.
When I returned to France, I joined this team in order to prepare
my thesis under the direction of Bill Triggs. And it was a natural
occurrence that, after six months, I began work on creating a start-up.
The idea was to find a new application based on the research we were
engaged in. The various prizes we have received lead me to believe
that I was right to take on this undertaking: Milpix was an award
winner at the 2007 Concours national d’aide à la création
d’entreprise de technologies innovantes [National Competition
for Assistance to Innovative Technology Enterprise creation] in the
Emerging division. And, in April 2008, our project also received
the Talent de l'économie prize awarded by the Grenoble Chamber
of Commerce and Industry, as well as Capital Week's "Best Angels" (1) in the context of its search for financial partners.
What would make this technology attractive to investors?
Christophe Bourez: Using Milpix, one can track down an original
based on a touched-up version of a photo. We can also offer our clients
the option of finding redundant copies of images, which might interest
archiving services and photo libraries, in particular. One more use,
among many others, to be added to the two current applications of the
software. I think new fields will open up in terms of everything having
to do with handling image indexation. But one shouldn't push the dream
too far, and should know how to distinguish what is possible from what
is not when the undiscovered scientific realm is too great. I thus don't
believe that one can yet manage to recognize people using images like
those of faces. Not right now, in any case.
(1) An event that, over one week, brings together players in France's
entrepreneurship sector in growth management and R&D investment. |
|
Milpix in brief
 Supplier
of software solutions for visual search and press monitoring
services.
In August 2007, Christopher Bourez, created the Milpix company
based on software of the same name. The fruit of ten years of research
at INRIA [the French national institute for research in computer science
and control], this software serves to locate in one second an image
in a database containing 10 million images, and this can be done using
any processor on the market. As an additional innovation, this search
is not done by keyword but rather based on the image itself. This feature
is one that only this start-up offers at present and one that allows
it to target many potential sales markets. Set up as a public limited
company, this start-up has seven partners, including Cordelia Schmid,
Director of the INRIA project-team LEAR in
Grenoble, and Hervé Jégou,
Research Manager at INRIA Grenoble - Rhône-Alpes Research Centre,
who act as scientific consultants, and Christopher Bourez's father,
who has worked for a major industrial group and contributes his experience
in business management.
About the Technology
 The
Milpix algorithm performs a search based on the contents of the
image. Using a set of fragments that can be modified, a complete
image is found among millions of images stored in thousands of
computers or on websites. The technology recognizes, in real time,
images that contain the same geometric shapes. For example, if
you submit the image of a street, the search engine will stop at
every photo taken on that same street. It is also capable of finding
the original photo using a touched-up version, which could prove
to be useful for purposes of copyright. One application is already
in use by a market watch service designed for the sectors of advertising,
communication, media planning and marketing, but also for any company
wishing to find and track photos and advertisements emanating from
its various departments. Whereas existing solutions are at present
all manually operated, Milpix is fully automatic and offers the
possibility of surfing and performing text and image searches in
databases of the French press via a directory containing 250 entries.
It is associated with an advertising checking service (daily bulletins)
for images (advertisements, photos, logos) that can be downloaded
to an on-line account that allows you to create a customized alert
system at low cost. Another application is undergoing a feasibility
study for a major French industrial client. It provides identification
of objects via a search string using a database that indexes 100
million objects that can be compared with 160 million images. The
final product is expected to be out a year from now.
|
French
version