Inspired by Real Biology
In 1981, David Hubel and Torsten Wiesel were awarded a Nobel Prize
for their discoveries concerning information
processing in the visual system.
In this work, they showed for the
first time that the visual cortex in Macaque monkeys is highly
structured, with many different identifiable functions being
performed by the cells there.
Despite its exciting implications, Torsten and Wiesel's work has
received limited application in the field of Computer
Vision. The Foveola shape recognition engine is the first
technology to integrate a broad range of these cellular functions -
with ground-breaking results.
The Foveola
A very large part of the visual cortex in primates
is devoted to processing nerve signals coming from the fovea,
a tiny area near the centre of the eye's retina.
At the centre of the fovea is an even smaller region known as the
foveola, containing a few thousand cone receptors in a
tightly packed array.
When we view a scene, the tiny foveola region is
sequentially fixated on different points of interest. This
process seems to have great significance in interpreting our
visual surroundings.
New Ways of Seeing
Inspired by the primate visual cortex, the innovative Foveola shape
recognition engine uses a fast, simulated-cell process to generate
a specific low-dimensional "code" for any given simple shape.
Excitingly, the Foveola engine groups related shapes into general
classes very similar to those naturally used by humans:
shapes that we consider similar have Foveola codes that are related
numerically in straightforward ways.