திட்டப்பணி
சுருக்கம்
இன்று
அனனத்து
இயற்னை
பேரிடர்ைளுை்கும்
பமலாை
உலைம்
எதிர்கைாள்ளும்
மிைே்கேரிய
பிரச்சனன
பைாவிட்
-19
ஆகும்
.
ஓராண்டுை்கு
பமலாகியும்
இே்பிரச்னனை்ைான
தீர்வு
இன்னும்
கதானலவில்
உள்ளது
.
இருே்பினும்
,
உலை
சுைாதார
அனமே்பின்
அறிவுறுத்தலின்ேடி
கெடிே்னேை்
ைட்டுே்ேடுத்த
இன்னும்
சில
ெழிைள்
உள்ளன
.
அெர்ைளில்
சிலர்
முைமூடி
அணிந்து
சமூை
இனடகெளினய
ைனடபிடிை்கின்றனர்
.
கோதுை்
கூட்டத்திபலா
அல்லது
நிைழ்விபலா
முைமூடிைனளை்
ைண்டறிெபத
திட்டத்தின்
பநாை்ைமாகும்
.
பநாை்ைத்னத
அனடய
முைமூடி
அணிந்த
மற்றும்
முைமூடி
அணியாத
ஒரு
சிலரின்
ேடம்
உள்ளீட்டு
தரவுத்கதாகுே்ோைே்
ேயன்ேடுத்தே்ேடுகிறது
.
முன்
கசயலாை்ைம்
,
தரவு
கேருை்ைம்
,
ேயிற்சி
,
பசாதனன
மற்றும்
ேடே்
பிரிவு
ஆகியெற்னற
உள்ளடை்கிய
திட்டத்தின்
பநாை்ைத்னத
அனடெதில்
சில
கசயல்முனறைள்
உள்ளன
.
கசயல்முனறைளுை்குே்
பிறகு
,
மாஸ்ை்
R-CNN
அல்ைாரிதம்
உதவியுடன்
,
முைமூடி
அணிந்தெர்ைள்
மற்றும்
முைமூடி
அணியாதெர்ைளின்
உள்ளீட்டு
தரவுத்கதாகுே்பின்
ஒரு
பிரிை்ைே்ேட்ட
ேடத்னதே்
கேறுபொம்
.
பின்னர்
,
மாதிரியானது
கெே்பைமனரே்
ேயன்ேடுத்தி
கசயல்ேடுத்தே்ேடுகிறது
,
அங்கு
மை்ைள்
முைமூடி
அணிந்திருே்ேதன்
வினளொைவும்
,
முைமூடினய
அணியாமல்
இருே்ேதன்
வினளொைவும்
v
சதவீதத்தில்
துல்லியமாை
இருை்கும்
.
இந்த
திட்டத்தில்
,
கோது
இடங்ைளில்
முைமூடி
அணியாதெர்ைனள
ைண்டறியும்
மாதிரினய
உருொை்கி
ெருகிபறாம்
.
விமான
நினலயங்ைள்
,
ரயில்
நினலயங்ைள்
,
ேணியிடங்ைள்
,
ேள்ளிைள்
மற்றும்
பிற
கோது
இடங்ைளில்
கோதுே்
ோதுைாே்புை்ைான
ெழிைாட்டுதல்ைளுை்கு
இணங்குெனத
உறுதிகசய்யும்
ெனையில்
இந்தத்
திட்டம்
நிைழ்பநர
ேயன்ோடுைளுடன்
இனணை்ைே்ேடலாம்
.
பைாவிட்
-19
கநருை்ைடியின்
மத்தியில்
,
பைாவிட்
-19
ேரவுெனதை்
ைட்டுே்ேடுத்த
உலை
சுைாதார
அனமே்பு
ெழங்கிய
ெழிைாட்டுதல்ைனள
அனடெதில்
இந்தத்
திட்டம்
ைெனம்
கசலுத்துகிறது
.
vi
ACKNOWLEDGEMENT
I thank the Lord Almighty, whose showers of blessings have made this
project a reality.
I would like to express my sincere thanks and deep sense of gratitude to my
guide,
Dr
.
M
.
Vijiyalakshmi
, Associate Professor, Department of Information
Science and Technology. She has been a constant source of inspiration and I thank
her providing me with the necessary counsel and direction to help me complete
this project.
My sincere thanks to,
Dr
.
S
.
Sridhar
, Professor and Head, Department of
Information Science and Technology for his kind support and for providing
necessary facilities to carry out the work and prepare the thesis.
I wish to record my sincere thanks to the members of review panel,
Dr.S.Sridhar
, Professor and Head,
Dr.N.Thangaraj
, Assistant Professor,
Dr.L.SaiRamesh
, Teaching Fellow and
Dr.Tina Esther Trueman,
Teaching
Fellow
,
Department of Information Science and Technology for their valuable
suggestions and critical reviews throughout the project.
JANANI A
vii
TABLE OF CONTENTS
CHAPTER NO TITLE PAGE NO
ABSTRACT
iii
ABSTRACT (TAMIL)
iv
LIST OF FIGURES
x
LIST OF ABBREVIATIONS
xi
1
INTRODUCTION
1
1.1
OVERVIEW
1
1.2
OBJECTIVE
1
1.3
PROBLEM STATEMENT
1
1.4
PROJECT DESCRIPTION
2
1.5
MOTIVATION
3
1.6
ORGANIZATION OF THE PROJECT
3
2
LITERATURE SURVEY
4
2.1 SUMMARY OF THE SURVEY
8
3
SYSTEM DESIGN
9
3.1
EXISTING SYSTEM
9
3.2
PROPOSED SYSTEM
9
viii
3.3
HARWARE REQUIREMENT
10
3.4
SOFTWARE REQUIREMENTS
10
3.5
SYSTEM ARCHITECTURE
10
3.6 DETAILED DESIGN
11
3.7
MODULE DESCRIPTION
12
3.7.1 Input images
12
3.7.2 Image Preprocessing
12
3.7.3 Data Augmentation
13
3.7.4 Model Training
14
3.7.5 Testing the Model
15
3.7.6 Image Segmentation using
19
Mask R-CNN
3.7.7 Implementing the Model in Opencv
21
4
IMPLEMENTATION AND RESULTS
22
4.1 IMAGE PREPROCESSING
22
4.2 DATA AUGMENTATION
23
4.3 MODEL TRAINING
24
4.4 TESTING THE MODEL
26
4.7 IMAGE SEGMENTATION USING
27
MASK R-CNN
ix
4.8 IMPLEMENTING THE MODEL IN
29
OPENCV
5
CONCLUSION AND FUTURE WORK
30
5.1 CONCLUSION
30
5.2 FUTURE WORK
31
REFERENCES
32
x
LIST OF FIGURES
FIGURE NO
FIGURE NAME
PAGE NO
3.5
System Architecture
11
3.6
Detailed Design
11
3.7
Convolutional Neural Network
15
3.8
Mask R-CNN
20
4.1
Output of Image Preprocessing
23
4.2
Output of Data Augmentation
24
4.3
Model Summary
25
4.4
Model Trained for 30 Epochs
26
4.5
Training and Validation Accuracy
27
4.6
Training and Validation Loss
27
4.7
Classification Report
27
4.8
Output of Image Segmentation
28
4.9
Face Mask Detection
29
xi
LIST OF ABBREVIATIONS
ABBREVIATION DESCRIPTION
WHO
World Health Organization
CNN
Convolutional Neural Network
MASK R-CNN
Mask Regional Convolutional Neural Network
FASTER R-CNN
Faster Regional Convolutional Neural Network
1
CHAPTER 1
INTRODUCTION
1.1 OVERVIEW
Coronavirus disease 2019 (COVID-19) is a contagious disease caused
by severe acute respiratory syndrome coronavirus. The disease is mainly
spread through the airborne route where the virus is transmitted via droplets
coming from infected people like cold, cough or sneeze etc.
The WHO has imposed various restrictions to control the spread of this
virus which includes wearing face masks and maintaining social distance. In
the middle of such an intense crisis this application can be used in social
gathering venues to detect the face masks that can help reduce the spread of
Covid-19.
1.2 OBJECTIVE
The objective of the project is to develop a model which can be used to
detect face masks in a group of people to prevent Covid-19. It is created using
MobileNet V2.
1.3 PROBLEM STATEMENT
In the middle of Covid-19 crisis, wearing masks is a fundamental need
nowadays. In public places because of the large volume of people it becomes
tough for security officials to check every person who is not wearing a mask.
This model detects a single person wearing a mask or not. Implementing a
2
similar principle to evaluate a group of people wearing masks is the real
challenge.
1.4 PROJECT DESCRIPTION
In this project, using MobileNet V2 algorithm to detect face masks. An
image with few people wearing a mask and not wearing a mask is sent as input
dataset and the segmented image of the same is obtained as output.
The first step is pre-processing where the image is resized to a particular
resolution and it is converted to a numpy array. Then, one hot encoding is
performed on the images. Finally the data set is split into training and testing.
The second step is data augmentation where the image is flipped,
zoomed and finally flipped horizontally.
The third step is training, where the base model is loaded with imageNet
weights and the last layer of the pre-trained model is fine tuned. In the last
fully connected layer, the average pooling, flattening, dense, activation
function (relu and softmax) and dropout value are entered and processed. Then
the model summary is obtained and configuration is saved.
In the fourth step, the optimizer loss entropy and accuracy metric are
configured and the trained model is evaluated and saved.
In the fifth step, the training loss and training accuracy data are plotted.
Next, a prediction on the training set is done. Finally, the model is evaluated.
The sixth step is image segmentation where the dataset is loaded and ID
mapping is done where mask and no mask is assigned. Next a tensorflow
session is created and the Mask RCNN model is loaded. Then, actual detection
3
of Boxes, Class, Scores and Masks are done. Then, instance segmentation is
performed and Detection results are visualized.
Finally the model is implemented using a webcam where the video is
read by frame and resized as necessary. Then, the preprocessing function is
called to get the result of people wearing a mask and not wearing a mask along
with the accuracy in percentage.
1.5 MOTIVATION
For the past 2 years, because of the deadly virus Covid-19, the entire
world has taken a spin from its routine. The spread of covid is still evident in
most places with new variants of the same virus. One thing that has become
common while going out nowadays is the face mask. But still there are many
people unaware of the situation and refusing / avoiding to wear a face mask.
This project helps in identifying people who are not wearing a face mask.
With the one of the common things added like wallet or hand bags, face mask
has become the latest one added.
Do'stlaringiz bilan baham: |