Information Technology Reference
In-Depth Information
faster evolutionary approach, although the resulting scene may not fully satisfy all
the constraints (which in scenes with more elements may actually be desirable).
To supplement the constraint-based and evolutionary approaches, we wanted the
teaching interface to enable the user to simply draw an example of the scene or
scene element that they wanted to specify, and for the software to use this as an
exemplar in order to generate similar looking examples. To do this, we implemented
a drawing interface that records key anchor points of each line drawn by a user. The
anchor points are recorded as variables rather than fixed values, so they can vary
within ranges in order to produce similar looking shapes in a scene. Additionally, we
allow the user to specify the hue, saturation and brightness ranges within which the
colour of each shape can vary, and to specify allowable linear transformations (such
as translations and rotations) and non-linear transformations (such as perspective
warping) that entire shapes, or even the entire scene can be subjected to.
To further supplement the scene generation abilities of the teaching interface, we
integrated the CFDG generative art software,
2
and our own evolutionary art software
(Hull and Colton
2007
, Colton et al.
2011
). The former system is able to generate
representational and abstract artworks by using context free design grammars, and
there are thousands of grammars available for use in art projects. The latter system
is able to generate abstract art forms in a number of styles, including pixel-based
(similar to fractals), particle based, and spirograph based (Colton and Browne
2009
).
Finally, the sixth scene generation method available within the teaching interface is
to take a digital image and turn it into a segmentation, as described in the non-
photorealistic rendering section above. We further enabled the image to be filtered
before it is segmented, as per our
Filter Feast
software (Torres et al.
2008
).
In addition to a screen for each of the segmentation generation methods described
above, the teaching interface has a screen to describe how the different methods are
to be used in layers to form the overall scene. It also has a screen to describe how
the different elements of the scene are to be rendered using NPR techniques and a
screen to describe how to generate paint dance animations (see below). The teaching
interface is currently in beta development. While it is not yet ready for general usage,
it is possible to define and render scenes. Given that we hope to attract other people
to use the software to define their own pictures, it was important to provide example
projects that produce interesting images. This has led us to the production of a series
of galleries, collectively entitled: “Ever so Slightly...”. The rather strange name
recognises the fact that it is not feasible that anyone will project a great deal of
imagination onto software which is able to produce novel scenes using a template
provided by people, but it may be possible to project a slight amount of imagination
onto the software, and this is our aim.
There are currently four galleries in the series, named: “PresidENTS”, “Fish
Fingers”, “After AARON”, and “Dance Floor”. An example from the “Dance
Floor” series has been given in Fig.
1.1
, and we give examples from the oth-
ers in Fig.
1.10
. The titles of the first two are fairly awful puns which reflect
2
Available at
www.contextfreeart.org
.
