Welcome to :
The FACE-IT project
by
A
Face-It Screen Example
Have
a look: Interactive
Demo it's available here !!
Hei
!!!!Downloadable
Demo it's available Now !!
Visitors
since 3rd of May 1996:
The
Face-it Project (FIP) package uses a Genetic Algorithm to evolve face expression
pictures and is meant to be used in psycological studies.The evolution
of pictures is based on the user's evaluation of a number of face expressions
shown on the screen.
The
FIP model originates from an inspiration of Leonard
Da Vinci Face Studies
and Artificial Life tecniques of Evolving Images, plus Cognitive Science
Studies of face Expressions Recognition.
A
Genetic Algorithm works on a population of genotypes. In FIP, the genotype
of each individual is represented by a bit string codifying the values
of 33 continue variables that describe 400.000 possible expressions of
the same face.
FIP
constructs a whole population of individuals (i.e.: nine). Initially, the
genotype of each individual will be set randomly. Each individual builds
a face that is shown on the screen in front of the user. According to the
genotype different faces will appear. The user selects same of the expressions
for riproduction by clicking on them with the mouse.Each selected individual
(i.e.: expression) will be cloned a fixed number of times and a new generation
of faces will appear on the screen. Each clone will be mutated in some
randomly chosen parts of the genotype (bit string). The average number
of mutations can be decided by the user. The selection and cloning process
can continue until one or more satisfactory expressions are obtained.
NOTES: We inform the reader that the Face-It Project is in
constant evolution towards a sharper definition of the algorithm and the
face expression images. Moreover, the FIP is now being tested to observe
people reactions and used to obtain scientific results about its application.
This
Project has been realized with the collaboration of:
This page is available on WWW since the 13th of September 1995.
Updated on the 2nd October 1995.
For
more information contact:
Luigi Pagliarini>
E-Mail:
[email protected] Go
back to Luigi Pagliarini Homepage