com.anji.floatingeye
contains classes to build and operate a
"floating eye" which is an agent that can move up and down, left and right, and zoom
in and out relative to an image, or "surface". The artificial retina is a 2x2 array
of pixels that is updated based on the state of the surface and the position of the
retina.
The central class of this package is coma.anji.FloatingEye
. It is meant
to identify an image as containing some target feature (e.g., its purpose might be to
identify a square when presented with several different shapes). It outputs an
affinity, a double
code> between 0 and 1 identifying level of confidence
that the image contains the target feature.
A floating eye's input values are stored in com.anji.nn.Neuron
objects.
A floating eye does not necessarily need to be controlled via an artificial neural
network (see com.anji.floatingeye.test.FloatingEyeTestDriver
for a
non-ANN example), but it was built primarily to be integrated with a
com.anji.nn.AnjiNet
object via
com.anji.floatingeye.AnjiNetFloatingEye
.