Vous êtes sur la page 1sur 27

An implementation of Artificial Neuron Network

Prepared By:

Fatima Nowshin Nayeema


ID: 072835050

Fahmida Khanam
ID: 083250050
Supervised By:
PrOnabananda Paul Course Instructor:
073764050 Dr. M. Rokonuzzaman
Overview
►Introduction
♦Human Face Recognition
♦Concept
›Computer Vision
› Artificial neural network

► Development Platform
► Diagram of face Recognition System
► Network Diagram
► Software
♦ Steps
♦Features Vector
♦Features
Overview
► Software
♦ Steps
♦Features Vector
♦Features
►Collection of features
►Software Simulation
►Building a model
♦Steps from 1-6
►Testing
►Performance
►Limitation
►Future Target
Human Face Recognition

▄ The face is the most distinctive and widely used key to


a person’s identity
▄The area of Face detection has attracted considerable
attention in the advancement of human-machine
interaction as it provides a natural and efficient way to
communicate between humans and machines.
▄This may seem like an easy task but the human face is a
dynamic object and has a high degree of variability in its
appearance,
▄ So, we will have to be smart to the each step of this
project.
Concept:

▄ Human Face Recognition system stand on two parts of


Computer Science:
1. Computer Vision
2. Artificial neural network

Computer Vision:
# It is used for localizing the object from the image.
# To find the edges for measurement.
# to make the proper illumination by applying mask.

◘In our project, we have not dealt with the computer vision. All the
tasks of it performed by image processing applying Software.
◘ Otherwise we become detached from our concentration. We
have insisted on the Neural Network part mostly.
Concept
Artificial Neural network (ANN):
◘ our main concentration is on the ANN.

◘An artificial Neural Network is a computational model capable of


computing various complex operations based on the concept of
biological neuron.

◘An artificial neural network learns what it should do from the learning
spaces after that it is tested for its performance .

◘After learning we run an ANN for the getting related output from it for
corresponding input.

◘ Learning is performed by various method.

◘ As, for this project we have used the ANN simulator, we don’t concern
with the learning algorithm.
Development Platform

◘ At each step we have taken help from the various


software.

Predict Domo: provided by Neuralware

◘ It is applied for creating Artificial Neural Network.


As, it is a demo version, we were unable to get full utility of it.

◘ Demonstration Software can analyze: 512 Records


23 Fields

GIMP: utility software of Ubuntu


For measurement of point to point distance

Photoshop 7: Adobe
For calculating pixel value and others preprocessing
Diagram of face Recognition
System
Network Diagram
Steps we have followed:

◘ Collection of Images:
We have collected images from a studio consisting One camera
man as there is a probability of being all the images of the
same type.
◘ after collecting the images, we have select the best 50 from
them to make the learning vector error free.
◘ Preprocessing those vectors by Photoshop
◘ Resize the images of the same type
Size:352X480 pixel

After that we have concentrated to find out the features vectors.


Features Vector
Features are:
• Breadth of Right Eye: BRE.........
• Height of Right Eye: HRE.........
• Breadth of Left Eye: BLE.........
• Height of Left Eye: HLE.........
• Length of Right Eyebrow: REB.........
• Length of Left Eyebrow: LEB.........
• Inner breadth of Nose: IBN.........
• Length of nose: LNO.........
• Distance of nose to Lip: DNLP........
• Length if lip: LLP.........
• Height of Lower Lip: HLP.........
• Breadth at Lip Zone: BLZ.........
• Face Breadth at Ear height: FBEH........
• Height of Forehead: HFH.........
• Value of Pixel from Nose Area: PXL-R........
• PXL-G........
• PXL-V........
Collection of features
• ◘ After selecting the image vectors, we take information from
each images.
• ◘ this measurement is taken by GIMP softeare.
• ◘ As, it it not done systemically, all the measurement are not
perfect at all.
• ◘ But we have tried our best to measure the point to point
distance.
• ◘ as, the used software performed on Excel file, we need to
input those information on the excel file.
• ◘ After building and learning the network, we test of the
network according to the guide of the software.
• ◘ and check it for the further output.
Software Simulation
To simulate the software
Open the Excell data file
From predict menu, select
the new then proceed the
following.
Building a model(Step1 0f 6)
Building a model(Step2 0f 6)
Building a model(Step3 0f 6)
Building a model(Step4 0f 6)
Building a model(Step5 0f 6)
Building a model(Step6 0f 6)
Testing
• After building the model, we have run the
test part of the soft.
Performance
• Testing is performed by 4 steps:
Partitioning data
– Required time: .77s

 Analyzing fields
– Required time: 17.43s

 Selecting variables and training


– Required time: 43.07s

Evaluating Model(50 Records)


– Required time: .73s
Performance:
• After training the network, when we train
for the input vectors for output, the we get
that only 34 out of 50 can be recognized
with 100% accuracy with relative entropy
• .0490. and the test is performed by 68.0%
accuracy with relative entropy .0333.
Future Target

◘ This project can be converted to an automated


system
◘By applying computer vision we can take the
measurement correctly. Then we will get more
precise network
◘ Being smart to select the image features, learning
accuracy can be improved.
Limitations
►This software can not retrieve information
directly.
►It can not recognize all the objects
accurately.
►It can recognize 34 objects accurately out
of 50 objects.
Thank you

Vous aimerez peut-être aussi