BackPropagation Neural Network for image Recognition by Ramesh

download BackPropagation Neural Network for image Recognition by Ramesh

of 38

Transcript of BackPropagation Neural Network for image Recognition by Ramesh

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    1/38

    BackPropagation Neural Network for

    image Recogn it ion

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    2/38

    Indira Gandhi Center For Atomic Research (IGCAR), is the

    second largest establishment of the atomic energy next to Bhabha

    Atomic Research Center, was setup at kalpakkam, 80 Kms south of

    Chennai in 1971 with the main objective of conducting Board

    based multidisciplinary programme of scientific research and

    advanced engineering, directed towards the development of sodium

    cooled Fast Breeder Reactor (FBR) technology, in India.

    ORGANIZATION PROFILE

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    3/38

    ABSTRACT

    Artificial Neural Networks are recent development tools that aremodeled from biological neural networks.

    The powerful side of this new tool is its ability to solve problems thatare very hard to be solved by traditional computing methods (e.g. byalgorithms).

    This work briefly explains Artificial Neural Networks and their

    applications, describing how to implement a simple ANN for imagerecognition.

    Back propagation, or propagation of error, is a common method of

    teaching Artificial Neural Networks how to perform a given task.

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    4/38

    Neural Network: Neural network is a computer program that can

    recognize the patterns in a given collection of data,and produce modelfor that data.

    Conventional computers use algorithmicapproach, if the specific stepsthat the computer needs to follow are not known, the computer cannot

    solve the problem.

    That means, traditional computing methods can only solve theproblems that we have already understood and know how to solve.

    However, ANNs are, in some way, much more powerful because theycan solve problems that we do not exactly know how to solve.

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    5/38

    That's why, of late, their usage is spreading over a wide range of

    areaincluding,

    robot control

    pattern (image, fingerprint, noise...) recognition.

    virus detection.

    Back Propagation ANNs contain one or more layers each of which arelinked to the next layer.

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    6/38

    The Existing System

    Consider an text processing task such as recognizing an everydayobject projected against a background of other objects. This is a taskthat even a small child's brain can solve in a few tenths of a second.

    But building a conventional serial machine to perform as well isincredibly complex. However, that same child might NOT be capableof calculating 2+2=4, while the serial machine solves it in a few

    nanoseconds. Ex: Hand-written characters also.

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    7/38

    Proposed System

    The attraction of neural networks is that they are best suited to solvingthe problems that are the most difficult to solve by traditionalcomputational methods.

    A large amount of input/outputdata is available, but you are not sure

    how to relate it to the output.

    The problem appears to have overwhelming complexity, but there isclearly a solution.

    The solution to the problem may change over time, within the boundsof the given input and output parameters.

    The output displayed as non-numeric.

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    8/38

    Software Specification

    Operating System Server: Windows XP Tools: Microsoft Visual Studio .Net-2008

    User Interface: Windows Application

    Code Behind: C#.Net 3.5

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    9/38

    Hardware Specification

    Processor: Intel Pentium

    Ram: 512 MB

    Hard Disk: PC with 40GB

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    10/38

    MODULES INVOLVED

    Load Network

    Train Network

    Recognize Topology of a Multi-Layer Perception

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    11/38

    LOAD NETWORK In this module a user can upload any image using this system

    user interface.

    By using this load network we can load a network file Network

    File (.net) extension.

    Here we are using one sample.net for load a network

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    12/38

    TRAIN NETWORK

    Apply input to the network.

    Calculate the output.

    Compare the resulting output with the desired output for the giveninput. This is called the error.

    Modify the weights for all neurons using the error.

    Repeat the process until the error reaches an acceptable value (e.g.error < 1%)

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    13/38

    RECOGNIZE

    This section describes a simple demonstration of the neural networklibrary, using the character classifier which comes with the source

    code.

    The network contains35 input nodes (one for each pixel)

    60 hidden nodes and

    26 output nodes

    The first layer is called the "input layer" which meets the initial input

    (e.g. pixels from a letter)

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    14/38

    and so does the last one "output layer" which usually holds the input's

    identifier (e.g. name of the input letter).

    The layers between input and output layers are called "hiddenlayer(s)" which only propagate the previous layer's outputs to the nextlayer and [back] propagates the following layer's error to the previous

    layer

    When in use (i.e. after training), the network will be loaded with acorrupted image, with a pixel value going to each input node.

    An alternative scheme would have 7 output nodes to give the binaryASCII character which best matches the given input.

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    15/38

    Topology of Multi-Layer Perception

    The principal importance of a neural network is not only the way aneuron is implemented but also how their interconnections (morecommonly called topology) are made.

    The topology of a human brain is too complicated to be used as amodel because a brain is made of hundreds of billions of connectionswhich can't be effectively described using such a low-level (andhighly simplified) model.

    The topology we will study is therefore not the topology of a humanbrain but actually a simple topology designed for easyimplementation on a digital computer.

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    16/38

    One of the easiest forms of this topology at the moment is

    made of three layers

    input layer (the inputs of our network)

    hidden layer output layer (the outputs of our network)

    All neurons from one layer are connected to all neurons in the

    next layer

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    17/38

    U Di

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    18/38

    Use-case Diagram

    Input Text

    Load Network

    Train Network

    Recognize The

    image

    Save Network &

    Layers Settings

    Admin User

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    19/38

    Data Flow Diagram

    Dataflow is the movement of data in a system from a point of

    origin to a specified destination indicated by line or arrow.

    Dataflow diagram is the graphical representation of the data

    movements, processes and files used in support of information

    systems.

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    20/38

    Context Level (0thlevel DFD)

    ADMIN

    Administrator

    Neural

    Network

    Image

    Recognition

    Data Input StageData

    Storage

    UI Screens

    Data Out put Stage

    Back PropagationUsing Neural Network

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    21/38

    Admin Functionalities1stLevel

    Open Form

    1.0.0

    Draw theImage

    1.0.1

    Load NeuralNetwork

    1.0.2

    Train theNetwork

    1.0.3

    Recongnize

    1.0.4

    Local Disk

    View Input

    and Output

    1.0.5Matched andUnmatched

    Heights Check

    1.0.6

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    22/38

    Admin Functionalities 2ndLevel

    Open Form

    1.0.0

    Choose the

    Existing

    Image

    1.0.1

    Browse theImage

    1.0.2

    Load theNetwork

    1.0.3

    Train the

    Network

    1.0.4

    Local Disk

    Recongnize

    1.0.5View Matched

    and Unmatched

    Heights Check

    1.0.6

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    23/38

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    24/38

    Home Page

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    25/38

    Drawing Own Text for Recognize

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    26/38

    Load Network

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    27/38

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    28/38

    Train Network

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    29/38

    Recognize The Text

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    30/38

    Save Network

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    31/38

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    32/38

    Choose an Existing Text for

    Recognize

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    33/38

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    34/38

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    35/38

    Providing Settings

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    36/38

    Save Settings

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    37/38

    It has been a great pleasure for me to work on this exciting andchallenging project. This project proved good for me as it

    provided practical knowledge of not only programming in

    ASP.NET and VB.NET web based application and no someextent Windows Application and SQL Server, but also about allhandling procedure related with Back propagation neuralnetwork for image recognition.

    It also provides knowledge about the latest technology used in

    developing web enabled application and client servertechnology that will be great demand in future. This will

    provide better opportunities and guidance in future indeveloping projects independently.

    Conclusion

  • 8/12/2019 BackPropagation Neural Network for image Recognition by Ramesh

    38/38