Unknown 16/10/23, 12:04 PM

Face Mask Detection

Introduction:

Since the end of 2019, infectious coronavirus disease (COVID-19) has been reported
for the first time in Wuhan, and it has become a public damage fitness issue in China and
even worldwide. This pandemic has devastating effects on societies and economies around
the world causing a global health crisis. It is an emerging respiratory infectious disease
caused by Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2). All over the
world, especially in the third wave, COVID-19 has been a significant healthcare challenge.
Many shutdowns in different industries have been caused by this pandemic. In addition,
many sectors such as maintenance projects and infrastructure construction have not been
suspended owing to their significant effect on people’s routine life [2].

Object detection is a computer vision approach for detecting and locating objects in
images and video. Object detection can detect instances of visual objects of specific classes
such as persons, animals, cars or building in digital pictures like photos or video frames.
When an image is shown to the eyes, the brain immediately recognizes the objects. On the
other hand, a machine needs long time and lot of training data to recognize the images.
However, due to the development of recent technology in machine learning, deep learning
and the field of computer vision has become a lot easier and more intuitive. Object detection
technology has exploded in popularity across a wide range of sectors. It enables self-driving
cars in safely navigating traffic, detecting violent behaviour in crowded areas, monitoring
object through video surveillance, recognising face mask through object detection module,
ensuring adequate quality control of pars in production, among many other things. And this is
only the tip of the iceberg in terms of what object detection technology can achieve!. In
modern video surveillance systems, detecting face in video streams is critical task. Deep
learning algorithms have recently been developed that deliver face identifying results [3].

Artificial intelligence enables image and video-based detection algorithms that can
accurately detect an object and determine whether the human is wearing or not wearing a
mask. Face mask identification can be done with diverse datasets utilizing deep learning and
machine learning approaches such as support vector machines and decision trees. The main
purpose of this thesis is to develop a Face mask detection model. The model was trained
using the MobileNetV2 architecture. Using two separate image datasets, the model was
trained and tested. A pretrained model of the MobileNetV2 architecture was used to
distinguish faces from video streams. In addition to the OpenCV framework, a variety of
packages of machine learning, deep learning approaches and image processing techniques
were used. To provide effective monitoring and enable proactively, the following processes
will be used: data augmentation, loading the classifier, establishing the fully connected layer,
pre-processing, and loading the picture data, applying the classifier, training phase, validation
and testing phase [1].

Objectives:

blob:https://web.whatsapp.com/c8b725a6-b22b-455c-bea7-40c696fb0a1c 1
Unknown 16/10/23, 12:04 PM
1. To investigate the effectiveness of Convolutional Neural Networks (CNNs) for real-
time face mask detection and recognition.

2. To develop a real-time, GUI-based automatic face detection and recognition system.

3. To assess the performance of the system in detecting and recognizing faces with and
without masks.

4. To examine the practical applications of the system in various fields, including

healthcare, security, and entry management.

Literature Survey:

1. “Face Mask Detection and COVID-19 Detection System”

blob:https://web.whatsapp.com/c8b725a6-b22b-455c-bea7-40c696fb0a1c 2
Unknown 16/10/23, 12:04 PM
• Authors: Mohammed Ghazy and Mohammad Kamel
• Published in: IEEE Access, 2020
• Summary: This paper discusses the integration of face mask detection with COVID-19
detection, using deep learning techniques to identify individuals without masks in real-
time. The system aims to contribute to controlling the spread of the virus in public
areas.

2. “Real-Time Mask Detection in Video Sequences”

• Authors: Ondrej Hubacek, Roman Oresnik, and Ludek Muller
• Published in: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition
Workshops (CVPRW), 2020
• Summary: This research focuses on real-time mask detection in video sequences using
deep learning, specifically a combination of YOLO (You Only Look Once) and
MobileNet. The proposed method is effective in identifying individuals wearing
masks[4].

3. “COVID-19: Detecting Face Mask Based on Deep Learning Approach”

• Authors: Aayush Tyagi, Dhanajay Yadav, and Mohit
• Published in: 2020 4th International Conference on Control, Automation and Robotics
(ICCAR), 2020
• Summary: This study explores the application of deep learning models, including
MobileNet and VGG16, to detect face masks in real-time. The authors analyze the
effectiveness of different model architectures and preprocessing techniques

4. “Real-time Face Mask Detection Using Deep Learning”

• Authors: Srinivas Reddy, Navneet and Raksha
• Published in: International Journal of Computer Science and Information Technology
Research, 2020
• Summary: This research presents a real-time face mask detection system that uses the
MobileNetV2 deep learning architecture. It discusses the system’s accuracy and its use
in various domains, including healthcare and security.

Methodology:

Various methods are used for face mask detection that’s include traditional techniques
like Haar cascades, HOG and SVM, as well as modern deep learning methods such as CNNs,
YOLO, Mobile Net-SSD, and transfer learning. These methods can be applied to images or
video streams and are used to identify whether a person is wearing a mask or not. The choice
of method depends on factors like accuracy and real-time processing requirements.

blob:https://web.whatsapp.com/c8b725a6-b22b-455c-bea7-40c696fb0a1c 3
Unknown 16/10/23, 12:04 PM

fig.1.Basic face mask detection system[5].

1.Support Vector Machine:

Support Vector Machines (SVM) play a role in face mask detection by training a
model to classify whether a face is wearing a mask or not. This process begins with a labelled
dataset containing two classes: faces with masks and faces without masks. Feature vectors are
extracted from the images to represent their content, capturing characteristics that can
differentiate between masked and unmasked faces. The SVM algorithm, a supervised
learning approach, then determines the optimal hyperplane, which is a boundary that
maximizes the margin between these two classes. When faced with a new, unseen image, the
SVM uses these features to classify it as either having a mask or not. While SVMs can be
effective for this task, deep learning methods have gained prominence due to their ability to
automatically learn relevant features, often achieving higher accuracy in face mask detection
applications.

2. Convolutional Neural Networks:

Using Convolutional Neural Networks (CNNs) for face mask detection is a robust and
widely adopted approach in computer vision. This method entails training neural networks to
recognize faces and discern whether they are wearing masks. Popular CNN architectures,
such as VGG and ResNet, are often employed for this task. A labeled dataset comprising
images of both masked and unmasked faces is essential for training the CNN. Through
training, the network learns to extract features that distinguish between the two classes, and it
can subsequently detect and classify faces in new images as either wearing a mask or not.
Transfer learning, leveraging pre-trained models, is common to boost performance, especially
when data is limited. Deep learning with CNNs has proven to be highly accurate in face mask
detection, making it a go-to choice for a wide array of real-world applications.

3. MobileNet-SSD:
MobileNet-SSD is a lightweight deep learning model designed for real-time
applications, making it an excellent choice for tasks like face mask detection where swift
responses are crucial. This model's efficiency is derived from its architecture, which balances
accuracy and speed. MobileNet-SSD can be fine-tuned to detect faces and masks by training
it on datasets containing labeled examples of both masked and unmasked faces. Once

blob:https://web.whatsapp.com/c8b725a6-b22b-455c-bea7-40c696fb0a1c 4
Unknown 16/10/23, 12:04 PM
adapted, it becomes a valuable tool for applications that require real-time monitoring of
individuals' compliance with mask-wearing mandates, such as in public places, healthcare
settings, or security systems. Its lightweight nature and real-time capabilities make it a
practical solution for various scenarios where both accuracy and speed are paramount.

4.YOLO (You Only Look Once):

YOLO (You Only Look Once) is an exceptionally efficient real-time object detection
algorithm in computer vision. Notably, its versions, such as YOLOv4 and YOLOv5, have
proven to be highly capable in the realm of face mask detection. What distinguishes YOLO is
its ability to simultaneously identify and locate objects within images and video frames,
making it ideal for applications that demand rapid and precise detection. YOLOv4 and
YOLOv5, being more recent iterations, bring even greater performance and accuracy. When
applied to face mask detection, these models are trained on datasets encompassing both
masked and unmasked faces, allowing them to swiftly recognize faces and categorize
individuals as mask-wearers or non-mask-wearers. This adaptability and real-time processing
make YOLO-based models a compelling choice for a wide array of face mask detection
applications, especially in scenarios where quick decision-making is essential for safety [4].

Convolution Neural Networks (CNN) Algorithm

Exploring the realm of face mask recognition, this study delves into a deep learning
algorithm featuring Convolutional Neural Networks (CNN) for classification. Specifically
crafted for pixel interpretation, CNNs excel in image recognition and analysis. Each layer of
the network employs a unique set of filters, ranging from hundreds to thousands, collectively
contributing to the final output. The cascade of results from one layer to the next
characterizes the neural network's progression. Evaluation of the proposed framework centers
on a face mask detection algorithm utilizing the TensorFlow software library, illustrated in
Figure 2. The Mask detector model is trained using Keras in tandem with TensorFlow. The
algorithm's procedural steps are delineated as follows.

blob:https://web.whatsapp.com/c8b725a6-b22b-455c-bea7-40c696fb0a1c 5
Unknown 16/10/23, 12:04 PM

Fig 2. Face Mask Detection[5]

STEP 1: DATASET COLLECTION

STEP 2: PRE-PROCESSING
STEP 3: SPLITTING
STEP 4: TRAINING
STEP 5: TESTING/EVALUATION

According to the above-mentioned algorithm, all the required dataset and components for
building the network is collected from various categories. Once the initial dataset is ready, the
next step is to train and test the set. This test dataset is used only in evaluating the
performance of the network. Next training should be done, so the neural network learns to
identify different categories in the given labels. Finally, the dataset should be evaluated and
compared with the ground-truth labels [5].

Plan of Research:
1. Literature Survey.
2. Study of related work on face mask detection using CNN deep learning.
3. Develop a real-time, GUI-based automatic face detection and recognition system.
4. To improve face mask detection using CNN deep learning.
5. Conclude the result of deep learning for face mask detection.

blob:https://web.whatsapp.com/c8b725a6-b22b-455c-bea7-40c696fb0a1c 6
Unknown 16/10/23, 12:04 PM

Implications:
The research focuses on the development of a real-time face mask detection system
using Convolutional Neural Networks (CNN). Trained on a carefully curated dataset
designed to simulate diverse real-world challenges, the CNN demonstrates the ability to
accurately identify face masks instantaneously. The system's significance lies in its
adaptability, navigating complex scenarios with varying lighting conditions, facial
orientations, and mask types.
This real-time capability holds implications for sectors such as public health, where it
can enforce mask-wearing protocols in high-traffic areas, and healthcare facilities, ensuring
safety compliance. Beyond health, the system finds utility in security and surveillance
applications, contributing to enhanced security measures and effective crowd management.
The research, by emphasizing real-time detection in diverse scenarios, positions the
CNN-based model as a versatile technology with potential societal impacts. Its nuanced
capabilities make it not only relevant for the current pandemic but also a preparedness
measure for future health crises. In essence, this research lays the groundwork for advanced,
adaptive systems in face mask detection that can contribute to broader societal well-being.

blob:https://web.whatsapp.com/c8b725a6-b22b-455c-bea7-40c696fb0a1c 7
Unknown 16/10/23, 12:04 PM

References:

[1]. Shah Nazir, Safa Teboulbi, Seifeddine Messaoud,” Real-Time Implementation of AI-Based
Face Mask Detection for COVID-19 Prevention”. Volume 2021 | Article ID 8340779.
[2]. Ondrej Hubacek, Roman Oresnik, and Ludek Muller,”“Real-Time Mask Detection in
Video Sequences”, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition
Workshops (CVPRW), 2020.
[3]. Srinivas Reddy, Navneet and Raksha,“Real-time Face Mask Detection Using Deep
Learning”,International Journal of Computer Science and Information Technology Research,
2020.

[4]. Mohammed Ghazy and Mohammad Kamel,“COVID-19: Detecting Face Mask Based on
Deep Learning Approach”,2020 4th International Conference on Control, Automation and
Robotics (ICCAR), 2020.
[5]. Hiten goyal, karanveer sidana, Charanjeet singh, Abhilasha jain, swati Jindal,”A real time
face mask detection system using CNN”,25 February 2022.
2

blob:https://web.whatsapp.com/c8b725a6-b22b-455c-bea7-40c696fb0a1c 8