Sie sind auf Seite 1von 9

EXPRESSION EVALUATION

USING ARTIFICIAL NEURAL NETWORK


SPECIFIC OBJECTIVE BEING
ADDRESSED BY THE PROJECT :

To implement a Artificial Neural Network(ANN)


for evaluating the expression
SCOPE OF THE PROJECT:

As the population and industrial sector continue to grow, so


does the amount of handwritten data that needs to be
processed. The expenses associated with paying human
employees to sort through the data and enter the
information manually into a computer becomes
increasingly more time-consuming and expensive. Knowing
this, it is easy to see that it would be beneficial to automate
part or all of the process.
TECHNOLOGICAL AIDS :

The MNIST database (Mixed National Institute of


Standards and Technology database) :Its a large
database of handwritten digits that is commonly used
for training various image processing systems.

Tensor flow
PROJECT BACKGROUND AND
LITERATURE REVIEW:
There exists a fair amount of research in this area. ANN forms the basis of our
experiment.

When used for image recognition, artificial neural networks (ANNs) consist of
multiple layers of small neuron collections which process portions of the
input image, called receptive fields. The outputs of these collections are then
tiled so that their input regions overlap, to obtain a better representation of
the original image; this is repeated for every such layer. Tiling allows ANNs to
tolerate translation of the input image.

The idea is to convert each image representing a digit into horizontal and
vertical histograms representing the number of filled in pixels in each row or
column, and used the resulting data as input to their genetic algorithm. The
goal of the genetic algorithm is to find a program or formula that could
distinguish numerals.
BLOCK DIAGRAM OF THE PROJECT
CHALLENGES :

One of the major problems encountered when


developing and comparing systems for digit recognition
is compiling a large standardized test sample database.
Many character recognition systems are based on the
concept of machine learning and require fine-tuning
parameters by processing training data.

Therefore, the test sample database must be large


enough to accommodate the amount of training these
systems require to tune such parameters.
CHALLENGES :

Another problem is that no one writes the same digit in


the exact same way. Intuitively, there is the problem of
handling the differences in rotation, skew, and relative
thickness of the edges of the digit written by two
different people. These distortions manifest as a result
of different writing styles, writing pressure, and other
factors. To account for these distortions, the pattern
recognition system should be trained to identify
common features shared between test samples that
represent the same digit.

Das könnte Ihnen auch gefallen