Professional Documents
Culture Documents
Batch 38
Batch 38
A PROJECT REPORT
Submitted by
PRIYADARSHINI A (312417104071)
of
BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE AND ENGINEERING
CHENNAI - 119
JULY-2021
ANNA UNIVERSITY : CHENNAI 600 025
BONAFIDE CERTIFICATE
SIGNATURE SIGNATURE
i
ACKNOWLEDGEMENT
We also take this opportunity to thank our honourable Chairman Dr. B. Babu
Manoharan, M.A., M.B.A., Ph.D. for the guidance he offered during our tenure in
this institution.
We express our deep gratitude to our honourable CEO Mr. B. Sashi Sekar,
M.Sc (INTL.Business) for the constant guidance and support for our project.
Our earnest gratitude to our Head of the Department Dr. J. Dafni Rose, M.E.,
Ph.D., for her commendable support and encouragement for the completion of the
project with perfection.
We wish to convey our sincere thanks to all the teaching and non-teaching
staff of the department of COMPUTER SCIENCE AND ENGINEERING without
whose co-operation this venture would not have been a success.
ii
CERTIFICATE OF EVALUATION
The reports of the project work submitted by the above students in partial
iii
ABSTRACT
COVID-19 pandemic caused by novel coronavirus is continuously spreading until now all
over the world. The impact of COVID-19 has been fallen on almost all sectors of
development. The healthcare system is going through a crisis. Many precautionary measures
have been taken to reduce the spread of this disease where wearing a mask is one of them. In
this paper, we propose a system that restrict the growth of COVID-19 by finding out people
who are not wearing any facial mask in a smart city network where all the public places are
monitored with Closed-Circuit Television (CCTV) cameras. While a person without a mask is
detected, the corresponding person and the administrator gets an alert. A deep learning
architecture is trained on a dataset that consists of images of people with and without masks
collected from various sources. The trained architecture achieved 81% accuracy on
distinguishing people with and without a facial mask for previously unseen test data. It is
hoped that our study would be a useful tool to reduce the spread of this communicable disease
iv
TABLE OF CONTENTS
CHAPTER TITLE PAGE
NO NUMBER
ABSTRACT iv
LIST OF FIGURES viii
1. INTRODUCTION 1
1.1 Overview 1
1.2 Problem Statement 2
1.3 Existing System 3
1.3.1 CNN Algorithm for Classification 3
1.3.2 YOLO used for Object Detection 4
1.4 Proposed System 4
1.4.1 Data Acquisition and Annotation 5
1.4.2 RESNET 50 for Feature Extraction 5
1.4.3 YOLO v2 for Object Detection and 5
Classification
2. LITERATURE SURVEY 7
3. SYSTEM DESIGN 12
3.1 Unified Modelling Language 12
3.1.1 Use Case Diagram of Face Mask 12
Detection System with automated door access
control 14
3.1.2 Class Diagram of Face Mask
Detection System with automated door access
control
3.1.3 Sequence Diagram of Face Mask 15
Detection System with automated door access
control
v
3.1.4 Activity Diagram of Face Mask 16
Detection System with automated door access
control
3.1.5 Collaboration Diagram of Face Mask 17
Detection System with automated door access
control
3.1.6 Component Diagram of Face Mask 18
Detection System with automated door access
control
3.1.7 Deployment Diagram of Face Mask 19
Detection System with automated door access
control
3.1.8 Package Diagram of Face Mask 19
Detection System with automated door access
control
4. SYSTEM ARCHITECTURE 21
4.1 Architectural Description 21
4.2 User Module 22
4.3 Server Module 22
4.3.1 Image Augmentation 23
4.3.2 Feature Extraction 23
4.3.3 Object Detection 23
4.3.4 Classification 23
5. SYSTEM IMPLEMENTATION 24
5.1 Implementation of Face Mask Detection 24
System
5.2 Pseudo code for Face Mask Detection System 26
vi
6. CODING AND SCREENSHOT 27
6.1 Face Mask Detection System using CNN 27
6.2 Output and Comparison 45
6.2.1 Training The Model 45
6.2.2 Results 45
6.2.3 Comparison 48
vii
LIST OF FIGURES
LIST OF NAME OF THE FIGURE PAGE NO
FIGURES
3.1 Use Case Diagram of Face Mask Detection System 13
with automated door access control
3.2 Class Diagram for Face Mask Detection System with 14
automated door access control
3.3 Sequence Diagram of Face Mask Detection System 15
with automated door access control
3.4 Activity Diagram of Face Mask Detection System 16
with automated door access control
3.5 Collaboration Diagram of Face Mask Detection 18
System with automated door access control
3.6 Component Diagram of Face Mask Detection System 18
with automated door access control
3.7 Deployment Diagram of Face Mask Detection 19
System with automated door access control
3.8 Package Diagram of Face Mask Detection System 20
with automated door access control
4.1 System Architecture for Face Mask Detection 21
System with automated door access control
viii
CHAPTER 1
INTRODUCTION
1.1 OVERVIEW
The coronavirus pandemic is defining global health crisis of our time and
the greatest challenge we have faced since World Wat Two. Since its emergence
in Asia in 2019, the virus has spread to every continent except Antarctic.
We have now reached the tragic milestone of more than two million
deaths, and the human family is suffering under an almost intolerable burden of
loss.
The situation now has been under attack and also growing badly in all
over the countries proclaimed by the WHO. According to this epidemic beyond
114 countries being affected by this flu and Million numbers of people getting
sick in one day.
During this disaster time period everyone should raise awareness and
naturally should do some oneself activities. By this issue the country’s
government, social authority and working place should strictly follow necessary
1
rules through continuous measurement and protest people’s health.
We want to build a system that can detect faces in real-world videos and
identify if the detected faces are wearing masks or not. If you look at the people
in videos captured by CCTV cameras, you can see that the faces are small, blurry
and of low resolution. People are not looking straight to the camera and the face
angles vary from time to time. These real-world videos are entirely different
from the videos captured by webcams or selfie cameras, making the face mask
detection problem much more difficult in practice.
In this system, we are going to detect the image of a person as input and
by using Yolo v2 deep learning algorithm we are going to classify them
accordingly. Also we are going to notify the person by sending alert message to
their mobile.
2
1.3 EXISTING SYSTEM
In the first system the learning model is based on CNN which is very
useful for pattern recognition from images. The network comprises an input
layer, several hidden layers and an output layer. The hidden layers consist of
multiple convolution layers that learn suitable filters for important feature
extraction from the given samples. The features extracted by CNN are used by
multiple dense neural networks for classification purposes.
3
1.3.2 YOLO USED FOR OBJECT DETECTION
In the second system YOLO model is used for creating bounding boxes
and object detection. An input here is an image is passed into the YOLO model.
This object detector is going through the image and find the coordinates that
present in an image. It basically divides the input into a grid and from that gird
it will analyze the target object’s features. From the neighboring cells that
features were detected with high confidence rate are add at one place for produce
model output.
4
This is also applicable for the staff members and the workers in the
campus. If any of the staff members is not wearing a mask then an alert is sent
to that corresponding person’s phone stating them to wear a mask. Also the
administrator will also get an alert, the image and the details of the employee
who is not wearing a mask to take necessary actions.
An input here is an image which passed into the YOLO v2 model. This
object detector is going through the image and find the coordinates that present
in an image. It basically divides the input into a grid and from that gird it will
analyze the target objects features. From the neighboring cells the features that
5
were detected with high confidence rate are added at one place for produce
model output. Finally bounding boxes are created and the image is classified as
‘with mask’ or ‘without mask’.
6
CHAPTER 2
LITERATURE SURVEY
7
Declared by the WHO that a potential speech by maintaining distance and
wearing a mask is necessary. Wearing a mask captured by the image detection
where the machine can cover and translate only the mouth portion of the face
part. Computer vision is a following section of Deep learning particularly an
area of convolution neural network (CNN). Added with one main thing is CNN
supports very high configuration Graphic Processing Units (GPU) thus as real
time image or video extraction of visualization is a bitter task. As we require
people mask having or not which call a surveillance system there is a need for
powerful validation such as video stream analysis that is fulfilled by advanced
CNN. Now cloud system real time video or image illustration among R-CNN
getting complex within create unsatisfied occurrence. Accordingly, new phase
included which have consisted 27 CNN layer across 24 convolutional layers that
fully connected algorithm is YOLO (You Only Look Once) as it affordable to
identified tiny objects. YOLO darknet algorithm while introduced to researchers
who covered pretty much different segments of image detection easily specially
face detection entirely grabbed YOLO.
8
sizes of the input and the output vectors were 298 and 76 respectively. Suitable
ANN structures and training algorithms were selected, parameters of them were
determined. ANN based IFPSF was trained to generate face masks more
realistically until achieving certain accuracy in learning. ANN based system
tested using 40 test sets covering only fingerprints.
Truong Quang Vinh and Nguyen Tran Ngoc Anh proposed a system
called Real -Time face mask detector using Yolo v3 algorithm and Haar Cascade
Classifier. This paper proposed a real-time face mask detection system using
Yolo v3 algorithm. In order to increase the processing time and accuracy of the
detector ,Haar cascade detector is used to detect the face region in input images
and then region of interest(ROI) is put into the Yolo v3 to detect the face mask.
The deep learning model has been trained with dataset of 7000 samples. Finally
a whole system was built to demonstrate an application in which people are
checked whether they are wearing face mask or not at the entrance of the door.
Susanto, Febri Alwan Putra, Riska Analia, Ika Karlina Laila Nur proposed
a system called The face mask detection for preventing the spread of Covid-19
at Politenik Negeri Batam. In this work, the object which needs to be detected is
the face mask wearer. The Yolo v4 which has been implemented in this work
consists of two-stages detector. The first-stage detector consists of input,
backbone, neck, and dense prediction. Moreover, the second stage of detector
has sparse prediction to predict the object by understanding the bounding boxes
and the classes on the object. The input image of this work needs to be handled
in the resolution of around 1920x1080 pixel in real-time application to ensure
the detection of the moving object who wear the face-mask properly. In this part,
10
the convolutional layer 3x3 is built and the larger number of parameters will be
selected as the backbone input layer. While at the backbone parts, the Darknet53
has been chosen as a detector method. In this backbone, the input network
resolution will be shrunk into 512x512 with receptive field size of 725x725 and
contains 29 convolutional layers by 3x3 and then each layer will be sent to the
neck detector. The PANet is applied as the neck detector method for
understanding the parameter aggregation from different backbone level detector.
All the aggregation layer from the neck detector will be sent to the sparse
prediction parts as the input layer in this stage. Meanwhile, the last detector
process on stage one is done at the dense prediction. The Yolo v3 model is used
in this stage to generate the prediction which result will be used as the input
prediction at the second-stage predictor. The second stage predictor has sparse
prediction which applies the faster R-CNN as the prediction method. In this
stage, it got the input layer from the neck which is 3x3 layer and the input
prediction from the dense prediction. The result from this stage are two classes
such as face-mask wearer and un-wearer of face-mask.
11
CHAPTER 3
SYSTEM DESIGN
In this chapter, the various UML diagrams for face mask detection system
with automated door access control is represented and the various functionalities
are explained.
Use case diagrams are considered for high level requirement analysis of a
system. So when the requirements of a system are analyzed the functionalities
are captured in use cases. So it can be said that uses cases are nothing but the
system functionalities written in an organized manner. Now the second things
which are relevant to the use cases are the actors. Actors can be defined as
something that interacts with the system. The actors can be human user, some
internal applications or may be some external applications. Use case diagrams
are used to gather the requirements of a system including internal and external
influences. These requirements are mostly design requirements. Hence, when a
12
system is analyzed to gather its functionalities, use cases are prepared and actors
are identified.
Figure 3.1 Use case diagram for face mask detection system with automated
door access control
The Functionalities are to be represented as a use case in the
representation. Each and every use case is a function in which the user or the
server can have the access on it. The names of the use cases are given in such a
way that the functionalities are preformed, because the main purpose of the
functionalities is to identify the requirements.
13
3.1.2 CLASS DIAGRAM FOR FACE MASK DETECTION SYSTEM WITH
AUTOMATED DOOR ACCESS CONTOL
Figure 3.2 Class diagram for face mask detection system with automated door
access control
The name of the class diagram should be meaningful to describe the aspect
of the system. Each element and their relationships should be identified in
advance responsibility (attributes and methods) of each class should be clearly
identified for each class minimum number of properties should be specified.
14
3.1.3 SEQUENCE DIAGRAM FOR FACE MASK DETECTION SYSTEM
WITH AUTOMATED DOOR ACCESS CONTOL
UML sequence diagrams model the flow of logic within the system in a
visual manner, enabling to both document and validate the logic, and are
commonly used for both analysis and design purposes.
Figure 3.3 Sequence diagram for face mask detection system with
automated door access control
The various actions that take place in the application in the correct
sequence are shown. Sequence diagrams are the most popular UML for dynamic
modeling.
15
3.1.4 ACTIVITY DIAGRAM FOR FACE MASK DETECTION WITH
AUTOMATED DOOR ACCESS CONTROL
Figure 3.4 Activity diagram for face mask detection system with automated
door access control
Activity diagrams are not only used for visualizing dynamic nature of a
system but they are also used to construct the executable system by using
forward and reverse engineering techniques. The only missing thing in activity
16
diagram is the message part.
Activity diagram is suitable for modeling the activity flow of the system.
It does not show any message flow from one activity to another. Activity
diagram is sometime considered as the flow chart. Although the diagrams looks
like a flow chart but it is not. It shows different flow like parallel, branched,
concurrent and single.
Now to choose between these two diagrams the main emphasis is given
on the type of requirement. If the time sequence is important then sequence
diagram is used and if organization is required then collaboration diagram is us
17
Figure 3.5 Collaboration diagram for face mask detection system with
automated door access control
Figure 3.6 Component diagram for face mask detection system with automated
door access control
18
3.1.7 DEPLOYMENT DIAGRAM FOR FACE MASK DETECTION WITH
AUTOMATED DOOR ACCESS CONTROL
Figure 3.7 Deployment diagram for face mask detection system with
automated door access control
19
collection of logically related UML elements. Packages are depicted as file
folders and can be used on any of the UML diagrams.
Figure 3.8 Package diagram for face mask detection system with automated
door access control
20
CHAPTER 4
SYSTEM ARCHITECTURE
In this chapter, the System Architecture for face mask detection system
with automated door access control is represented and the modules are
explained.
Figure 4.1 System Architecture for face mask detection system with automated
door access control
The major modules of this system are User module, Server module, Dataset,
RESNET 50 used for Feature Extraction, YOLO v2 for Object detection and
classification.
21
4.2 Interactions between User, Server and Dataset
In the user module, the image is captured through web camera and sent to
the server module for image augmentation and feature extraction and further
classified into classes using the previously trained datasets.
The image is given as input by the user or in real time image is captured
through web camera and further processing are done. Data Acquisition is
performed by the system from the user.
The system extracts data from the user in the form of image and performs
augmentation, feature extraction , object detection and classification. In
augmentation, size of the dataset is increased. In feature extraction, weights and
bias for input image is extracted. In object detection, bounding boxes are created
22
and object is detected and classified using algorithms.
4.3.4 CLASSIFICATION
23
CHAPTER 5
SYSTEM IMPLEMENTATION
In this chapter, the System Implementation for the Face Mask Detection
System model is explained in detail.
We proposed a face mask detection system to monitor people who are not
using a face mask. In this project, all public places such as schools, colleges and
offices are monitored by CCTV cameras or web cameras. The cameras are used
to capture images from public places; then these images are feed into a system
that identifies if any person without face mask appears in the image. If any
person without a face mask is detected then this information is sent to the
administrator to take necessary actions.
24
learning architecture, we collected images from different sources. The
architecture of the learning technique highly depends on YOLO v2. All the
aspects of deep learning architecture are described below.
1) Dataset Collection: Data from two different sources are collected for
training and testing the model. We collected a total of 100 images of people with
masks and without a mask. For training purposes, 70% images of each class are
used and the rest of the images are utilized for testing purposes.
START
TO DO SEGMENTATION
PRODUCE OUTPUT
SUGGEST PRECAUTIONS
STOP
26
CHAPTER 6
The following code corresponds to the face mask detection system using
YOLO v2. The input of the code is image of a person with or without mask. The
output is the class of the mask -> with mask/ without mask.
load COVID19_Mask_yolo.mat
I = imread("sample.jpg");
Ir = imresize(I, resz);
if(~isempty(bbox))
27
bboxf = bboxre(bbox, size(I), resz);
end
figure
clf
imshow(detectedImg)
bbox(:,1) = bbox(:,1)*sz(2)/targetSize(2);
bbox(:,2) = bbox(:,2)*sz(1)/targetSize(1);
bbox(:,3) = bbox(:,3)*sz(2)/targetSize(2);
bbox(:,4) = bbox(:,4)*sz(1)/targetSize(1);
end
load('COVID19_Mask_yolo.mat');
28
Load test movie
vidobj = VideoReader('sample.mp4');
viewer = vision.DeployableVideoPlayer;
currAxes = axes;
while(hasFrame(vidobj))
sz = size(frame);
if(~isempty(bbox))
end
step(viewer, detectedImg);
end
29
function bbox = bboxre(bbox, sz, targetSize)
bbox(:,1) = bbox(:,1)*sz(2)/targetSize(2);
bbox(:,2) = bbox(:,2)*sz(1)/targetSize(1);
bbox(:,3) = bbox(:,3)*sz(2)/targetSize(2);
bbox(:,4) = bbox(:,4)*sz(1)/targetSize(1);
end
cont = 1;
useMex = false;
executionEnvironment = 'cpu';
w = webcam();
w.Brightness = 64;
w.Resolution = '1280x720';
viewer = vision.DeployableVideoPlayer();
viewer.Size = "Full-screen";
Auxiliary variables
30
fps = 0;
avgfps = [];
% Handle response
switch answer
case 'SSD'
load COVID19_Mask_SSD.mat
mdl = 'SSD';
case 'YOLOv2'
load COVID19_Mask_yolo.mat
mdl = 'YOLOv2';
end
while cont
img = snapshot(w);
sz = size(img);
bbox = [];
31
tic;
if useMex
if strcmp(mdl, 'SSD')
label = label';
end
else
end
newt = toc;
detectedImg = img;
if ~isempty(bbox)
num = numel(bboxf(:,1));
else
end
viewer(detectedImg)
cont = isOpen(viewer);
end
release(viewer)
bbox(:,1) = bbox(:,1)*sz(2)/targetSize(2);
bbox(:,2) = bbox(:,2)*sz(1)/targetSize(1);
bbox(:,3) = bbox(:,3)*sz(2)/targetSize(2);
bbox(:,4) = bbox(:,4)*sz(1)/targetSize(1);
end
doLabeling = true;
33
if doLabeling
imageLabeler
else
load groundTruth.mat
end
Convert labeled ground truth from Image Labeler app into table data type.
trainingDataset = gTruth.LabelData;
trainingDataset.files = gTruth.DataSource.Source;
head(trainingDataset)
index = 1;
% Colorized
Iout = imread(trainingDataset.files{index});
for k = 1: width(trainingDataset)-1
bboxes = table2array(trainingDataset(index,k+1));
if ~isempty(bboxes{1})
34
trainingDataset.Properties.VariableNames{k+1},'Color','r','fontsize', 72,
'linewidth', 30);
end
end
figure, imshow(Iout);
rng(1004);
shuffledIndices = randperm(height(trainingDataset));
trainingData = trainingDataset(shuffledIndices(1:idx),:);
validationData = trainingDataset(shuffledIndices(validationIdx),:);
testData = trainingDataset(shuffledIndices(testIdx),:);
imdsTrain = imageDatastore(trainingData{:,'files'});
35
bldsTrain = boxLabelDatastore(trainingData(:,2:end));
imdsValidation = imageDatastore(validationData{:,'files'});
bldsValidation = boxLabelDatastore(validationData(:,2:end));
imdsTest = imageDatastore(testData{:,'files'});
bldsTest = boxLabelDatastore(testData(:,2:end));
trainingData = combine(imdsTrain,bldsTrain);
validationData = combine(imdsValidation,bldsValidation);
testData = combine(imdsTest,bldsTest);
basenetwork = resnet50();
numClasses = width(trainingDataset)-1;
Sweep over a range of values and plot the mean IoU versus number of
anchor boxes to measure the trade-off between number of anchors an mean IoU.
maxNumAnchors = 15;
36
meanIoU = zeros([maxNumAnchors,1]);
doTrain = true;
if doTrain
for k = 1:maxNumAnchors
[anchorBoxes{k},meanIoU(k)] =
estimateAnchorBoxes(anchorTraining,k);
end
else
load AnchorMask.mat
end
figure
plot(1:maxNumAnchors, meanIoU,'-o')
ylabel("Mean IoU")
xlabel("Number of Anchors")
grid on
numAnchors = 10;
allBoxes =
round(cell2mat(reshape(table2array(trainingDataset(:,2:end)),[],1)));
37
scale = inputSize(1:2)./size(imread(trainingDataset.files{1}), [1 2]);
anchorBoxes = round(anchorBoxes{numAnchors}.*scale)
meanIoU_chosen = meanIoU(numAnchors)
nettotrain = "YOLOv2";
if nettotrain == "YOLOv2"
featureLayer = 'activation_40_relu';
end
Data Augmentation
augmentedTrainingData = transform(trainingData,@augmentData);
Read the same image multiple times and display the augmented training
data.
augmentedData = cell(4,1);
for k = 1:4
38
data = read(augmentedTrainingData);
augmentedData{k} =
insertShape(data{1},'Rectangle',data{2},'linewidth', 15);
reset(augmentedTrainingData);
end
figure
preprocessedTrainingData =
transform(augmentedTrainingData,@(data)preprocessData(data,inputSize,
nettotrain));
preprocessedValidationData =
transform(validationData,@(data)preprocessData(data,inputSize, nettotrain));
options = trainingOptions('adam',...
'InitialLearnRate',0.001,...
'Verbose',true,...
'MiniBatchSize',8,...
'MaxEpochs',50,...
'Shuffle','never',...
'VerboseFrequency',10, ...
39
'ValidationData', preprocessedValidationData);
doTrain = true;
if doTrain
if nettotrain == 1
[detector,info] =
trainYOLOv2ObjectDetector(preprocessedTrainingData,lgraph,options);
elseif nettotrain == 2
[detector,info] =
trainSSDObjectDetector(preprocessedTrainingData,lgraph,options);
end
detector
You can verify the training accuracy by inspecting the training loss for
each iteration.
figure
plot(info.TrainingLoss)
grid on
xlabel('Number of Iterations')
else
if nettotrain == 1
40
load COVID19_Mask_yolo.mat
elseif nettotrain == 2
load COVID19_Mask_SSD.mat
end
end
I = read(testData);
I = imresize(I{1},1);
if(~isempty(bboxes))
end
figure
clf
imshow(I)
preprocessedTestData =
transform(testData,@(data)preprocessData(data,inputSize,nettotrain));
figure
plot(recall, precision)
xlabel('Recall')
ylabel('Precision')
grid on
cfg = coder.gpuConfig('mex');
cfg.TargetLang = 'C++';
42
cfg.DeepLearningConfig = coder.DeepLearningConfig('cudnn');
if nettotrain == 1
else
end
Helper Functions
if nettotrain == 2
data{1} = imresize(data{1},targetSize(1:2));
data{2} = bboxresize(data{2},scale);
else
data{1} = imresize(data{1},1);
data{2} = bboxresize(data{2},1);
end
end
function B = augmentData(A)
B = cell(size(A));
43
I = A{1};
sz = size(I);
I = jitterColorHSV(I,...
'Contrast',0.2,...
'Hue',0,...
'Saturation',0.1,...
'Brightness',0.2);
end
rout = affineOutputView(sz,tform,'BoundsStyle','CenterOutput');
B{1} = imwarp(I,tform,'OutputView',rout);
[B{2},indices] = bboxwarp(A{2},tform,rout,'OverlapThreshold',0.25);
B{3} = A{3}(indices);
if isempty(indices)
B = A;
end end
44
6.2 OUTPUT AND COMPARISON
The following figure is the training taking place in the model after
the augmentation steps. The model uses Yolo v2 for training. Once the data set
is loaded and following the data augmentation , feature extraction the figure
shows the model fitting taking place with 50 epochs.
6.2.2 RESULTS
The Algorithm and the page designs are implemented in the MATLAB
R2019a. Here, the various functionalities required for the applications are
implemented by coding them in Python. The various functionalities used here are
RESNET 50 and YOLO v2
For every end user who uses the Window, open the MATLAB using the run
command. Once the MATLAB is opened, copy the path of the Code Location and
open the particular coding module.
45
6.1 Opening the MATLAB
As a first process we will be copying the path of the code where we saved.
46
Output for still image:
47
6.2.3 COMPARISON
Displaying the learning rate, total loss, and the individual losses (box loss,
object loss and class loss) for every iteration. These can be used to interpret how
the respective losses are changing in each iteration. For example, a sudden spike
in the box loss after few iterations implies that there are Inf or NaNs in the
predictions
48
CHAPTER 7
7.1 CONCLUSION
This paper presents a system for a smart city to reduce the spread of
coronavirus by informing the authority about the person who is not wearing a facial
mask that is a precautionary measure of Covid-19. The motive of the work comes
from the people disobeying the rules that are mandatory to stop the spread of
coronavirus. The system contains a face mask detection architecture where a deep
learning algorithm is used to detect the mask on the face. To train the model, labeled
image data are used where the images were facial images with masks and without a
mask. The proposed system detects a face mask with an accuracy of 81.0%. The
decision of the classification network is transferred to the corresponding authority.
The system proposed in this study will act as a valuable tool to strictly impose the
use of a facial mask in public places for all people.
49
problem in the network. The proposed system mainly detects the face mask and
informs the corresponding authority with the location of a person not wearing a
mask. Based on this, the authority has to send their personnel to find out the person
and take necessary actions. But this manual scenario can be automated by using
drones and robot technology to take action instantly. Furthermore, people near to the
person not wearing a mask may be alerted by an alarm signal on that location, and
displaying the violators face in a LED screen to maintain a safe distance from the
person would be a further study.
50
REFERENCES
[7] L. Liu et al., “Deep Learning for Generic Object Detection: A Survey,”
Int. J. Comput. Vis., vol. 128, no. 2, pp. 261–318, Sep. 2018.
[14] J. Won Sonn and J. K. Lee, “The smart city as time-space cartographer
in COVID-19 control: the South Korean strategy and democratic control of
surveillance technology,” Eurasian Geogr. Econ., pp. 1– 11, May. 2020.
52