From even have hundred layers to discover different

From a single glance a human can understand the world which is assumed
to be a great accomplishment. All it takes for seconds to categorize certain
environment or an object and its importance in the field of scene recognition.
One of the human capability is to learn and memorize places by analyzing the
world in some seconds. Our neural design constantly saves different input even
for a short span of time. It is common that people can recognize scene they see
such as a mosque, tomb or any fort. The recent researches show that in 36 milliseconds
processing time with 80% accuracy viewer can recognize the scene. Now question
raises how much an artificial machine will learn before stretching out
possession of human being and how we can recognize scene that fast and what information
do we need?

It is important to
understand scene perception because in researches it if found that a scene uses
our certain knowledge that is connected to some scene category (eg, mosque, tomb,
fort). Such knowledge forces that we must pay attention and it may be helping
in recognizing a particular object and determines what kind of information do
we need to memorize from the scene. Researchers on scene recognition travel a
point between two system i.e cognition and perception and it is a problem that
it considered to be a challenging task for a person working in artificial intelligence.
This kind of researches can be used to design a system in artificial intelligence
that have the ability to recognize the scene and its category

A CNN is
a class of deep learning and feed forward AI that has been successfully applied
in scene recognition and visual imagery. Convolutional
neural network directly learn from image data set and removing the need to
manually feature extracting to produce better recognition result. CNN are much like ordinary neural and CNN consist of multilayers that
is used to minimal preprocessing. CNN is also known as shift variant. Conventional
neural network appears to be successful in many real life studies and
application such as image classification, face recognition and much more. We
will have to go back 2012 to understand its success when Alex Krizhevsky used CNN to
win 2012 imageNet event, resulting in error reduced from 26% to 15%

 

A convolutional neural network may even have
hundred layers to discover different features in an image. CNN architecture is design to take 2D
structure advantage. It is achieved by local connection and tied weight followed
with certain pooling that result the invariant translation feature. Additional advantage
of conventional neural network is that it is trained easily and require few
parameters that can connect fully operational network with identical no. of hidden
units. CNNs use a variation of multilayer
followed by full connected layer designed to require minimal preprocessing. Convolutional networks were inspired by biological processes in which the connectivity pattern
between neurons is inspired by the organization of the
animal visual
cortex. CNN was inspired
with biological process and it contains connection pattern between neurons which
are inspired by animal organization visual cortex. In CNN every neuron will be receiving
input and carry out a dot product followed by non-linearity