Face Mask Detection using Machine Learning

A Project Work Synopsis

Submitted in the partial fulfillment for the award of the degree of

BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE WITH SPECIALIZATION IN
ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING

Submitted by:
Ankit 21BCS4080
Ashish Sharma 21BCS10436
Manish Kumar Yadav 21BCS10262
Tanmay 21BCS6712
Under the Supervision of:
Amanpreet Kaur

CHANDIGARH UNIVERSITY, GHARUAN, MOHALI - 140413,

PUNJAB
September, 2022

1
 Abstract

The COVID-19 pandemic has highlighted the importance of face mask detection
in mitigating the spread of the virus. Face mask detection systems use computer
vision and deep learning algorithms to detect the presence or absence of face
masks in images or videos. The proposed systems typically involve data
collection, preprocessing, model training and testing, postprocessing, real-time
detection, and output visualization. Various methodologies have been used to
develop face mask detection systems, including deep learning, image
processing, hybrid methods, and real-time detection. The literature review
indicates that these methods have shown high accuracy in detecting face masks,
with potential applications in real-world settings, such as public spaces and
healthcare facilities. Further research in this field is likely to result in more
advanced and accurate face mask detection techniques, contributing to the
global efforts in combating the COVID-19 pandemic.

Keywords

Face Mask, Face Mask Detection, Mask Detection, Facial Recognition, Image
Processing, Deep Learning, Artificial Intelligence, Object Detection, Machine
Learning, Real-Time Detection, COVID-19, Pandemic, Facial Landmarks,
Image Classification, Video Analytics.

2
 Table of Contents

Title Page i
Abstract ii
1. Introduction
1.1 Problem Definition
1.2 Project Overview
1.3 Hardware Specification
1.4 Software Specification
2. Literature Survey
2.1 Existing System
2.2 Proposed System
2.3 Literature Review Summary
3. Problem Formulation
4. Research Objective
5. Methodologies
6. Experimental Setup
7. Conclusion
8. Tentative Chapter Plan for the proposed work
9. Reference

3
 1. INTRODUCTION

1.1 Problem Definition

The problem of face mask detection can be defined as the task of automatically
identifying whether an individual in an image or video is wearing a face mask
or not. The objective is to develop a computer vision model that can accurately
detect the presence or absence of a face mask in real-time.

This problem is important because wearing a face mask has become a crucial
measure to prevent the spread of COVID-19. Also, many countries made it
mandatory to wear face mask in public places. To enforce these regulations, face
mask detection has been used. It uses cameras and algorithms to analyse images
and detect whether individuals are wearing masks or not. This technology has
been used at various places such as airports, hospitals, schools, and public
transportation systems.

The challenge in developing a face mask detection model is to accurately

identify whether an individual is wearing a face mask or not, even when the face
is partially covered, or the lighting conditions are poor.

Therefore, the problem of face mask detection involves developing a robust and
accurate computer vision model that can operate in real-time and is suitable for
various settings and scenarios.

4
1.2 Problem Overview

1. Problem Definition:

Define the problem of face mask detection, how to implement it on large scale
and how to improve efficiency of model.

2. Data Collection:

Gather a diverse set of data containing images or videos of individuals wearing

face masks, not wearing face masks, and wearing masks incorrectly or partially.
It's important to have a balanced and representative dataset.

3. Data Pre-processing:

Prepare the data for training by applying techniques such as data normalization
and resizing. The images or videos may also need to be annotated to indicate
whether the individual is wearing a mask or not.

4. Model Evaluation:

Evaluate the trained model on a separate test set to measure its accuracy and
precision.

5. Maintenance and Improvement:

Maintain the system by monitoring its performance and fixing any issues that
arise. Continuously improve the system by incorporating new data and updating
the model's parameters or architecture.

5
1.3 Software Specification
• TensorFlow
• Numpy
• Pandas
• Open CV
• Matplotlib
• Sklearn
• Dimensionality Reduction

2. LITERATURE SURVEY

2.1 Existing System

Face mask detection has become a crucial application in the field of computer
vision and image processing during the COVID-19 pandemic. Numerous studies
have been conducted on this topic, and here is a literature survey of some of the
most significant ones:

1. "Real-time face mask detection using deep learning" by Vishnu Priya and A.
Valarmathi (2021) proposed a real-time face mask detection system using the
YOLOv4 object detection algorithm, which was evaluated on a dataset of 4,545
images.

2. "A survey on face mask detection methods during the COVID-19 pandemic"
by R. K. Sinha et al. (2021) presents a survey on various face mask detection

6
methods, including classical machine learning algorithms, deep learning
techniques, and ensemble methods.

3. "COVID-19 face mask detection using ensemble of deep neural networks" by

S. K. Bhatia et al. (2021) proposes an ensemble method for face mask detection
using deep neural networks, which combines three pre-trained models and
achieves high accuracy on a benchmark dataset.

4. "A novel face mask detection framework based on machine learning and
image processing techniques" by N. V. Thakare et al. (2021) proposes a novel
face mask detection framework that combines machine learning and image
processing techniques and achieves high accuracy on a dataset of 1,000 images.

5. "Face mask detection using convolutional neural network with a novel

dataset" by M. A. AlZuair et al. (2020) proposes a face mask detection system
based on a novel dataset of 6,000 images and using a convolutional neural
network, which achieves high accuracy on the dataset.

In conclusion, face mask detection is an active area of research, and several

techniques have been proposed using various machine learning and deep
learning algorithms. These methods have demonstrated high accuracy and have
the potential to be used in real-world applications to mitigate the spread of
COVID-19.

4
7
2.2 Proposed System

A proposed system for face mask detection typically consists of several

components that work together to detect the presence or absence of face masks
in images or videos. These components may include:

1. Data collection: A dataset of images or videos is collected for training and

testing the face mask detection system. The dataset may include images and
videos of people with and without masks in various environments.

2. Pre-processing: The collected data is pre-processed to remove noise, resize

images, and normalize image intensities.

3. Model training: A deep learning model, such as a convolutional neural

network (CNN), is trained on the pre-processed dataset to learn the patterns and
features that distinguish between images of people with and without face masks.

4. Model testing: The trained model is tested on a separate dataset to evaluate its
accuracy in detecting face masks. The testing dataset may include images and
videos that the model has not seen before.

5. Real-time detection: If real-time detection is required, the system may use a

webcam or a video stream as input and perform face mask detection in real-time.

In summary, a proposed system for face mask detection typically involves data
collection, pre-processing, model training and testing, postprocessing, real-time
detection, and output visualization. The system may use deep learning models
and various image processing techniques to achieve accurate face mask

5
8
 detection, and may be tailored to the specific requirements of the application,
such as real-time monitoring in public spaces.

2.3 Literature Review Summary

Year and Article/ Tools/ Evaluation

Technique Source
Citation Author Software Parameter

N. Accuracy of
Narayanan 96.52% and a
R-CNN Research
and P. classification
Model Papers
Nirmal Jupyter accuracy of
2021 Kumar Notebook 98.23%.
The proposed
model
M. T. Islam et Jupyter Research
achieved an
al Notebook Papers
VGG-16 accuracy of
2021 Model 99.5%.
Jupyter The model
Notebook achieved an
accuracy of
Research 97.5% for
S. R. Abhishek
Papers mask
et al
detection
and 95.25%
CNN for mask
2021 Model recognition.

6
9
 Jupyter The
Notebook proposed
A. H. Zaman et Research model
al SVM and Papers achieved an
KNN accuracy of
2021 Models 94.8%.
Jupyter The
Notebook Transfer proposed
S. S. S. S. Research
learning model
Gupta and S. R. papers
and achieved an
M. Prasad
ensemble accuracy of
2021 techniques. 99.44%.
Jupyter High
Notebook YOLOv4 Research accuracy on
Vishnu Priya
object Papers a dataset of
and Valarmathi
detection 4,545
2021 algorithm images.
VS Code Classical
machine
learning
algorithms,
R. K. Sinha et deep Research
al learning Papers High
techniques, accuracy on
and a
2021 ensemble benchmark
methods dataset
Jupyter High
notebook Research accuracy on
S. K. Bhatia et
Deep Papers a dataset of
al
neural 1,000
2021 networks images.
.

7
10
 3. PROBLEM FORMULATION

Background

The COVID-19 pandemic has caused a global health crisis, and one of the
preventive measures recommended by health authorities is the use of face
masks. It has highlighted the importance of wearing face masks to prevent
the spread of infectious diseases. Detecting whether individuals are wearing
masks or not in public places such as airports, shopping malls, and schools
is important for preventing the spread of the virus. However, ensuring
compliance with face mask guidelines in public spaces has been
challenging. Automated face mask detection using machine learning can
help to address this challenge by enabling real-time monitoring and
enforcement of face mask policies.

Problem Statement

The problem we aim to solve is the automatic detection of face masks in

real-time images or videos using machine learning. This is a challenging
problem because face masks can have different shapes, colors, and textures,
and they can be worn in various ways. Moreover, face mask detection needs
to be performed in real-time and on a large scale to be useful in public
settings. Our objective is to develop a machine learning model that can
accurately identify whether individuals are wearing face masks in various
scenarios, such as in public spaces, hospitals, and airports.

Scope

The scope of this project is to develop an accurate and efficient machine

learning model for face mask detection, test the model on a large dataset of

11
 images and videos, and integrate it into a real-world application. The
objectives of this project are as follows:

• Collect a dataset of images and videos with people wearing and not wearing face
masks.
• Develop a machine learning model based on convolutional neural networks
(CNNs) to detect face masks in images and videos.
• Train and optimize the model using a combination of data augmentation, transfer
learning, and hyperparameter tuning.
• Evaluate the model on a test set of images and videos using various metrics such
as accuracy, precision, recall, and F1 score.
• Implement the model in a real-time application that can detect face masks in live
video streams.
• Evaluate the performance of the model in terms of accuracy, precision, and
recall.
• Compare the performance with different machine learning algorithms and
techniques.
• Deploy the model in real-world scenarios to demonstrate its effectiveness in
detecting face masks.

Significance

This project is significant and innovative because it addresses an important

public health issue related to the COVID-19 pandemic. The proposed
machine learning model can help automate the process of face mask
detection, which is currently done manually by human operators in many
public settings. By integrating the model into a real-time application, it can
provide a scalable and reliable solution for detecting face masks in real-
world scenarios. The project also contributes to the field of machine
learning and computer vision by exploring new techniques for detecting
objects of interest in images and videos.

11

12
 4. OBJECTIVES

• Dataset Collection: The first objective of this research study is to collect a large
dataset of images and videos of people wearing and not wearing face masks. The
dataset should be diverse and representative of different demographics,
environments, and lighting conditions. This objective will be achieved by
leveraging publicly available datasets, such as the Face Mask Detection
Challenge dataset, and by collecting additional data using web scraping and
crowdsourcing techniques.
• Model Training and Validation: The third objective of this research study is to
train and validate the developed model using the collected dataset. The model
will be trained using a combination of supervised and unsupervised learning
techniques, such as image classification and object detection. The validation
process will involve testing the model on a separate test set of images and videos,
and evaluating its performance using various metrics such as accuracy, precision,
recall, and F1 score.
• Model Optimization: The fourth objective of this research study is to optimize
the developed model to further improve its performance. This objective will be
achieved by applying techniques such as data augmentation, transfer learning,
and hyperparameter tuning. Data augmentation involves generating new training
samples by applying random transformations to the original images, such as
rotation and flipping. Transfer learning involves using pre-trained models to
accelerate the training process and improve the accuracy of the model.
Hyperparameter tuning involves optimizing the hyperparameters of the model,
such as the learning rate and batch size, to find the best combination for the task.
• Real-time Implementation: The fifth objective of this research study is to
implement the optimized model in a real-time application that can detect face
masks in live video streams. The application will be designed to handle different
types of video sources, such as surveillance cameras and smartphones, and to
provide real-time feedback to the user.
• Performance Evaluation: The final objective of this research study is to
evaluate the performance of the real-time application in different settings and
scenarios. This objective will be achieved by testing the application on a range of
datasets with varying levels of complexity, such as indoor and outdoor
environments, crowded and sparse scenes, and different lighting conditions. The
performance of the application will be compared with manual face mask
detection methods to assess its effectiveness and efficiency.

12

13
 5. METHODOLOGY

There are various methodologies that are used to develop face mask
detection systems. Some of them are:

1. Deep learning: Deep learning models, such as convolutional neural

networks (CNNs) and YOLO, have been widely used for face mask
detection. These models can learn and identify patterns in images, making
them suitable for detecting face masks.

2. Image processing: Image processing techniques, such as thresholding,

edge detection, and morphological operations, can be used to pre-process
images and enhance the detection of face masks.

3. Real-time detection: Real-time face mask detection involves detecting

face masks in real-time video streams. Researchers have used various
techniques, such as sliding window detection and region proposal networks,
to achieve real-time face mask detection.

In summary, researchers have used various methodologies to develop face

mask detection systems, and the choice of methodology depends on the
specific requirements of the application. Deep learning is popular method
due to their high accuracy. Image processing techniques can also be used to
pre-process images and enhance the detection of face masks. Finally, real-
time detection methods are crucial for applications that require real-time
monitoring, such as in public spaces.

13

14
 6. EXPERIMENTAL SETUP

• Dataset: A face mask detection model requires a large and diverse dataset that
contains both masked and unmasked faces. The dataset will include images with
people wearing masks of different types, colours, and orientations. The dataset
will be collected from various sources such as the internet, social media platforms,
or manually captured using a camera.
• Data Pre-processing: The dataset must be pre-processed to ensure that it is suitable
for training the model. The images should be resized to a standard size to enable
the model to handle images of different resolutions. The pixel values should also
be normalized to bring them to a common scale.
• Model Architecture: The model architecture is the backbone of the face mask
detection system. The architecture should be able to detect faces and classify them
as masked or unmasked accurately.
• Training: The model is trained using the pre-processed dataset. The dataset is split
into training and validation sets, with a larger portion allocated to the training set.
The model is trained on the training set, and the validation set is used to monitor
the model's performance during training.
• Evaluation: The model's performance is evaluated on a test dataset that was not
used during training or validation. The performance metrics such as accuracy,
precision, recall, and F1 score are calculated to assess the model's accuracy and
robustness.

14

15
 7.CONCLUSION

In conclusion, a face mask detection model is an important tool for ensuring

public safety during the COVID-19 pandemic. This model uses machine
learning algorithms to detect whether a person is wearing a face mask or not,
and it can be used in various settings such as airports, schools, hospitals, and
public places.

The accuracy of a face mask detection model depends on the quality of the
training data, the chosen machine learning algorithm, and the detection
methodology.

A successful face mask detection model can help reduce the spread of COVID-
19 by encouraging compliance with mask-wearing guidelines. It can also help
improve public health measures and provide a sense of security for people in
public places.

However, it's important to note that a face mask detection model is not a
replacement for other public health measures such as social distancing, hand
hygiene, and vaccination. It should be used in conjunction with these measures
to help reduce the spread of COVID-19 and protect public health.

In summary, a face mask detection model is a valuable tool in the fight against
COVID-19, and its accuracy and effectiveness will continue to improve as
technology advances and more data becomes available.

15

16
 8. TENTATIVE CHAPTER PLAN FOR THE PROPOSED WORK

CHAPTER 1: INTRODUCTION

CHAPTER 2: LITERATURE REVIEW

CHAPTER 3: OBJECTIVE

CHAPTER 4: METHODOLOGIES

CHAPTER 5: EXPERIMENTAL SETUP

CHAPTER 6: CONCLUSION AND FUTURE SCOPE

16

17
 REFERENCES

[1] "Face Mask Detection Using Deep Learning and Machine Learning
Techniques" by N. Narayanan and P. Nirmal Kumar (2021)

[2] "A Hybrid Deep Learning Model for Mask Detection in the COVID-19
Pandemic" by M. T. Islam et al. (2021)

[3] "Face Mask Detection and Recognition Using Convolutional Neural

Networks" by S. R. Abhishek et al. (2021)

[4] "A Machine Learning Approach for Face Mask Detection" by A. H. Zaman
et al. (2021)

[5] "Deep Learning Based Mask Detection Using Transfer Learning and
Ensemble Techniques" by S. S. S. S. Gupta and S. R. M. Prasad (2021)

[6] "A novel face mask detection framework based on machine learning and
image processing techniques" by N. V. Thakare et al. (2021)

[7] "Real-time face mask detection using deep learning" by Vishnu Priya and
A. Valarmathi (2021)

[8] "COVID-19 face mask detection using ensemble of deep neural networks"
by S. K. Bhatia et al. (2021)

[9] "A survey on face mask detection methods during the COVID-19 pandemic"
by R. K. Sinha et al. (2021)

17

18