CSE4004 – DIGITAL FORENSICS

DIGITAL ASSINGMENT
NAME: AJAYJITH.N.S
REG.NO:18BCE2332
ABSTRACT
In recent times, where Covid-19 has impacted a domino effect on manufacturing, travel, tourism,
hospitality, crippling the global economy. In addition to it, is the growing curve of human deaths
across the globe due to the pandemic, this project which relies on computer vision and deep
learning, intends to make an impact and solve the real-world problem of safety measures at some
significant level.
Thus, there should be face mask detection in public places to identify who all are the ones that
violate the rule and regulations. This can be used at airports, offices, hospitals and many more
public places to ensure that the safety standards are maintained and people are abiding by the
rules and regulations to wear protective masks at public places. If the detection system classifies
as ‘No Mask’, reminders can be given as well as actions can be taken against such individuals.
INTRODUCTION
AIM
To improve life of the general society and capture the culprits who take advantages of the
paramedic and does crimes wearing face masks.
PROBLEM STATEMENT
In recent times everyone knows it is compulsory to wear face masks in public places so that the
spread of COVID-19 can be prevented. But some people take this opportunity to a misleading
advantageous side where they rob local store and other grocery shops. This research paper aims
to improve life of general society and capture these culprits who miss use the situation such a
this pandemic.
LITERATURE REVIEW
1. Detection of Face Mask in Real-time using Convolutional Neural Networks and Open-
CV

With the widespread transmission of COVID-19 in the year 2020, the world saw a
massive outcry. Wearing a face mask has become necessary as a preventative measure
against virus transmission, particularly in open situations where social distance is
frequently ineffective. As a result, wearing a face mask is critical for controlling the
pandemic's spread. Using convolutional neural networks (CNN) and Open-CV, this
research proposes a simplified method for detecting face masks in real-time. The
proposed CNN model uses two convolutional layers to train on a dataset of 12000 images
 of faces with and without mask, and the Haar cascade classifier from Open-CV to predict
on real-time video streams. Using the CNN architecture, the model achieved an accuracy
score of 98.8 percent on the training set and 99.37 percent on the validation set without
any overfitting issues.
2. Deep Learning Framework for Face Mask Detection

Our lives have been impacted by COVID-19, a pandemic sickness caused by the corona
virus. Since the epidemic, wearing face masks, cleaning, and maintaining social distance
have been the new normal. Because the Corona virus enters our bodies through the mouth
and nose, covering those places can assist to protect us from the infection. Despite the
fact that vaccines and treatments are available for this disease, WHO recommends that
masks be worn and that Covid protocols be followed. After the initial wave reached a
point of control and vaccination began, the populace began to disregard face masks and
social separation. As a result, the second wave has expanded widely, and several densely
populated countries are losing control. Wearing a mask will now greatly aid in the
prevention of disease transmission. This project aids with the detection of faces in live
video feeds, both with and without masks. The implementation is completed using
libraries such as TensorFlow, SciPy, OpenCV, and Keras. MTCNN is the face detection
model used. The overall goal of this project is to create an embedded model that can
detect faces in video streams and produce an accurate categorization of faces with and
without mask. This model can be used to conduct public surveillance and to require
people to wear masks in order to avoid the spread of the hazardous corona virus. The
classification accuracy of the suggested model is 97.2 percent.

3. Preventing Against Virus Transmission by Detecting Facemask for Gateway

Operation System

The COVID-19 epidemic forced governments all over the world to impose quarantines in
order to prevent the virus from spreading. Wearing a face mask at work, according to the
documentation, reduces the risk of spreading. AI is being used to provide a safe
environment in a production setup that is both efficient and cost-effective. A combined
model integrating machine learning and deep learning will be used to recognise face
masks. We'll use OpenCV to detect faces in real time from a live feed from our camera,
utilising a face mask detection library with photographs with and without masks. This
dataset is used to build a COVID-19 face mask detector with CV using OpenCV, Python,
Tensor Flow, and other technologies in this paper. The work's main goal is to use deep
learning and computer revelation to determine whether the person on the image/cinematic
clip is wearing a face mask or not. We will create a gateway system employing this face
mask detection. People can only enter this system if they wear a face mask. To build this
system, we used a Raspberry Pi and an a4899 driver module to operate the stepper motor.
The motor, which is attached to the driver module, controls the gateway.

4. Study of masked face detection approach in video analytics

 Because security is so important, video surveillance has become a hot focus of research.
By performing activities such as real-time incident detection and post-event analysis,
video analytics improve video surveillance systems. This can save human resources,
money, and improve the surveillance system's effectiveness. One of the most typical
security requirements for Video Analytics is to automatically recognise the presence of a
masked individual. We present a technique for masked face detection in this study that
involves four steps: estimating camera distance, eye line detection, facial part detection,
and eye detection. The ideas employed in each of these processes are outlined in the
paper, as well as the usage of widely accessible people detection and face detection
algorithms. This novel approach to the problem has resulted in a method that is simpler in
complexity, allowing for real-time implementation. The performance of the algorithm on
test video sequences provides significant information for future improvements in the
masked face detection performance.

5. Detecting Faces With Face Masks

The purpose of this research is to compare and contrast numerous methods for detecting
faces when they are hidden by a mask. As feature-based algorithms, Haar cascade and
Histogram of Oriented Gradients are examined, while convolutional neural network-
based approaches include Multitask Cascade Convolutional Neural Network, Max
Margin Object Detection, and TinyFace. Disposable face masks, burkas, balaclavas, ski
helmets with ski goggles, hockey helmets with protective grills, costumes, and others are
all considered. The TinyFace method achieves the highest accuracy, but it also
necessitates a lot more computing resources than other methods. As a result, this study
discusses an attempt to explore if retraining some of the surviving methods' models with
additional image data featuring faces with varied face masks can enhance their accuracy

6. COVID-19 Monitoring System using Social Distancing and Face Mask Detection on
Surveillance video datasets

COVID-19 virus is still a source of concern and hazard in today's world. With a huge
population moving around and an insufficient task force and resources to administer
them, manual monitoring of social distance standards is impossible. A lightweight,
durable, and 24-hour-a-day, seven-day-a-week video-monitoring system is required to
automate this operation. Using object detection, clustering, and a Convolution Neural
Network (CNN) based binary classifier, this study provides a comprehensive and
effective solution for person detection, social distancing violation detection, face
detection, and face mask classification. On surveillance video datasets, YOLOv3,
Density-based spatial clustering of applications with noise (DBSCAN), Dual Shot Face
Detector (DSFD), and MobileNetV2 based binary classifier were used. A comparison of
different face detection and face mask classification methods is also presented in this
work. Finally, to compensate for the lack of dataset in the community, a video dataset
 labeling method is presented, coupled with a labeled video dataset, which is utilized to
evaluate the system. The system's performance is measured in terms of accuracy, F1
score, and forecast time, which must be short in order to be practical. On the labeled
video dataset, the system has an accuracy of 91.2 percent and an F1 score of 90.79
percent, with an average forecast time of 7.12 seconds for 78 frames of video.

7. Image-based Face Detection and Recognition

In biometrics research, face recognition from an image or video is a hot topic.

Surveillance cameras for video recording are commonly found in public spaces, and these
cameras provide considerable security benefits. Face recognition is widely acknowledged
to have played an essential part in surveillance systems because it does not require the
cooperation of the object. Face-based identification has two distinct benefits over other
biometrics: distinctiveness and acceptance. Face detection is a difficult topic in computer
vision because the human face is a dynamic object with a great degree of diversity in
appearance. The precision and speed of identification are critical in this profession.

8. An improved face recognition algorithm and its application in attendance

management system

Face Recognition is a computer programme that detects, tracks, identifies, and verifies
human faces in images or videos acquired with a digital camera. Although much progress
has been achieved in the field of face detection and recognition for security, identity, and
attendance purposes, there are still challenges that are impeding progress toward human-
level accuracy. Variations in human facial appearance, such as variable lighting
conditions, noise in face photos, scale, attitude, and so on, are among these concerns.
This study proposes a new method that combines the Local Binary Pattern (LBP)
algorithm with advanced image processing techniques such as Contrast Adjustment,
Bilateral Filter, Histogram Equalization, and Image Blending to address some of the
issues that impede face recognition accuracy, thereby improving the LBP codes and thus
the overall face recognition system's accuracy. The results of our experiments reveal that
our method is very accurate, dependable, and resilient for a face recognition system that
can be used in a real-world setting as an autonomous attendance management system.

9. Study of Face Recognition Techniques

Face recognition has received a lot of attention in recent years, thanks to the increasing
rise of multimedia information. Face, as an object, has specific traits that must be
detected; as a result, it remains the most difficult research subject for computer vision and
image processing researchers. We attempted to handle the most difficult facial traits in
this survey work, such as posture invariance, aging, illuminations, and partial occlusion.
When applied to facial photographs, they are regarded as critical components in face
recognition systems. In addition, this paper examines state-of-the-art face detection
 techniques and approaches, such as Eigen face, Artificial Neural Networks (ANN),
Support Vector Machines (SVM), Principal Component Analysis (PCA), Independent
Component Analysis (ICA), Gabor Wavelets, Elastic Bunch Graph Matching, 3D
morphable Models, and Hidden Markov Models. In addition to the aforementioned
works, we have highlighted many testing face databases for results analysis, including
AT & T (ORL), AR, FERET, LFW, YTF, and Yale. However, the goal of this study is to
give a comprehensive assessment of the literature on face recognition and its
applications. After a thorough discussion, some of the most important findings are
presented in the conclusion.

10. A Survey paper for Face Recognition Technologies

Biometrics is the study of human behavior and characteristics. Face recognition is a

biometric approach. It is approached in a variety of ways. This paper includes a survey of
all of these strategies, as well as an analysis of various algorithms and methods. Face
recognition is a new field of biometrics for security, because no face can be defeated as a
security method. This study explains how we can recognise a face with the help of
computers.

11. Real-time facial feature detection using conditional regression forests

Despite the fact that facial feature identification from 2D photos is a well-studied field,
real-time approaches for estimating feature points, especially on low-quality images, are
lacking. For this objective, we suggest conditional regression forest. While regression
forests learn the relationships between facial image patches and feature point locations
across all faces, conditional regression forests learn the relationships that are conditional
on global face attributes. We demonstrate that conditional regression forests outperform
regression forests for facial feature recognition in our trials by using the head posture as a
global attribute.

12. Multi-Stage CNN Architecture for Face Mask Detection

Coronavirus Disease 2019 (COVID-19) emerged at the end of 2019, and it is still causing
havoc on the lives and businesses of millions of people in 2020. As the world recovers
from the pandemic and seeks to return to normalcy, there is an increase in anxiety among
those who plan to resume their everyday routines in person. According to various studies,
wearing a face mask minimizes the likelihood of viral transmission and offers a sense of
security. It is, however, impossible to carefully track the application of this policy.
Technology is the key here. A Convolutional Neural Network (CNN)-based architecture
for detecting instances of incorrect use of face masks is presented. Our system is
compatible with CCTV cameras and uses a two-stage CNN architecture that can
recognise both masked and unmasked faces. This will make it easier to track safety
infractions, promote the usage of face masks, and create a safe working environment.
13. Face Recognition based Attendance System using Haar Cascade and Local Binary
Pattern Histogram Algorithm

Face recognition based attendance systems are more secure and time-saving. There are
several research papers focusing on only the recognition rate of students. This research
focuses on a face recognition based attendance system with getting a less false-positive
rate using a threshold to confidence i.e. euclidean distance value while detecting
unknown persons and saving their images. Compared to other euclidean distance-based
algorithms like Eigenfaces and Fisherfaces, Local Binary Pattern Histogram (LBPH)
algorithm is better . We used Haar cascade for face detection because of their robustness
and LBPH algorithm for face recognition. It is robust against monotonic grayscale
transformations. Scenarios such as face recognition rate, false-positive rate for that and
false-positive rate with and without using a threshold in detecting unknown persons are
considered to evaluate our system. We got a face recognition rate of 77% and its false-
positive rate of 28%. This system recognizes students even when students are wearing
glasses or have grown a beard. Face Recognition of unknown persons is nearly 60% for
both with and without applying threshold value. Its false-positive rate is 14% and 30%
with and without applying threshold respectively.

14. Real Time Face Recognition Using LBP Features

Biometric systems for pattern recognition are in demand. Face Recognition has been a
subject of interest for researchers for years .In past few years,computational model
development has been difficult, but computing systems got help from advanced
technology to carry out face recognition. The computer applications emerged from semi-
automated models to accurate mathematical models. Analyzing geometry of faces
attained in representation of faces. The implementation of this in these area, with widen
use in industry solutions indicates the importance of the topic and its evolution. So, the
standard of face recognition technologies raised regularly, with implementation of more
accurate and faster systems. In this method, the face recognition is carried out by using
real time database which consists of various expressions and illumination conditions.
LBPH is utilized for the features extraction of the test image and KNN used for the
training.

15. Face Mask Detection Using AI

Covid-19 has created an entirely new frequency, and the people are becoming aware that
they are entering a new planet. While our society is rapidly changing, we must be nimble
in order to respond to new requirements that have ringed us all. Every person's thinking
will be focused on creating risk-free environments so that life can continue to be as
conductive as before. Options must be considered in order to protect all individuals
returning to our workplace and to keep ourselves and our loved ones safe. Every day,
 new strategies have been developed to comply with laws and regulations. Although face
masks are becoming a new standard for daily life, it is vital to be vigilant throughout the
day and to take action against individuals who do not wear masks in public places or at
work in order to create a safe environment that contributes to public safety. For the sake
of safety, many segments of society appear to be accepting of some Covid tracking tools.
Face mask detector is one of the most important pieces of equipment. This technology
can determine who is missing a mandatory face mask. These technologies use existing
surveillance systems and cutting-edge neural network algorithms to determine whether or
not a person is wearing a face mask. In this chapter, we'll go over artificial intelligence
and its subsets, such as machine learning and deep learning, as well as deep learning
frameworks and a simple face mask detection system implementation.

16. Face Mask Detection by using Optimistic Convolutional Neural Network

The COVID-19 epidemic has accelerated worldwide health concerns and is disrupting
our daily lives. Wearing a protective facemask to prevent the spread of coronavirus is one
of the reasons for survival advice. Wearing a facemask is the most efficient way to
protect yourself from COVID-19. It's tough to physically check if people are wearing
their facemasks appropriately and to warn the victim in public and crowded places. This
research proposes a simplified method for detecting and informing individuals who are
not wearing a facemask. The suggested system/model is trained and tested using Kaggle
datasets. The technology operates in real-time and identifies whether a person's face is
covered by a facemask. If it isn't, the person is notified individually via text message. The
mask is created

17. Real-Time Face Mask Detection Method Based on YOLOv3

COVID-19's fast spread has caused significant harm and infected tens of millions of
individuals around the world. Because there is no specific cure for COVID-19, wearing
masks has proven an effective technique of preventing transmission and is mandatory in
most public places, resulting in an increase in demand for automatic real-time mask
detection systems to replace manual reminders. However, there are only a few research
on face mask detection. Mask detector performance must be improved immediately. The
Properly Wearing Masked Face Detection Dataset (PWMFD) was proposed in this paper,
and it includes 9205 photos of mask-wearing samples divided into three groups. Squeeze
and Excitation (SE)-YOLOv3, a mask detector with roughly balanced effectiveness and
efficiency, was also proposed. By incorporating the SE block into Darknet53, we were
able to incorporate the attention mechanism and obtain the relationships between
channels, allowing the network to focus more on the relevant feature. To increase the
stability of bounding box regression, we used GIoUloss, which can better express the
spatial difference between predicted and ground truth boxes. The significant foreground-
background class imbalance was solved via focal loss.
 18. Control The COVID-19 Pandemic: Face Mask Detection Using Transfer Learning

Currently, the world is dealing with a health crisis caused by the Coronavirus COVID-19,
which has spread across the globe. For many countries, fighting the pandemic has become
an inevitable reality. Many fields of research in the application of new information
technologies, particularly those connected to artificial intelligence, are currently involved.
We provide a fresh contribution to the fight against the pandemic in this research. It is
about detecting persons who are wearing masks because they are unable to work or move
around normally without protection from COVID-19. However, there are just a few study
investigations on the detection of face masks. We studied the use of several deep
Convolutional Neural Networks (CNN) to extract deep characteristics from photos of faces
in this study. Various machine learning classifiers, such as Support Vector Machine (SVM)
and K-Nearest Neighbors, are used to further process the retrieved data (K-NN). To
compare the model performances, all different measures such as accuracy and precision
were employed and studied. The best classification rate was 97.1 percent, which was
obtained using a combination of SVM and the MobileNetV2 model. Despite the tiny
dataset (1376 photos) employed, we were able to achieve excellent results for the detection
of masks on faces.
CURRENT CHALLENGES

RELATED WORK:

➢ FebriEye, a thermal camera, comes with fresh analytics similar as face mask and social
distancing monitoring system which generates an alarm in case of any violations. It's
being developed by Vehant Technologies which is to be enforced by the Telangana
government.
➢ Uber has verified to CNN Business that it’s taking face masks or analogous coverings for
both motorists and passengers in countries like the US, and is developing technology to
descry whether or not motorists are abiding by those rules.
➢ Face Mask Alert app which is in development process by LeewayHertz software results.
It sends an alert to the druggies administering them to wear masks.
➢ AIZOOTech face mask discovery system which uses dataset composed of WIDER Face
and MAFA, but lacks corner net for the purpose of face alignment.
GAP ANALYSIS
➢ In recent times the Telangana government has implemented the thermal cameras which
can check the individuals temperature and if there is dramatic rise in the temperature then
the security system is alerted and the person will be asked leave the premises. For adding
a new feature where the face structure can be analyzed then the data can be stored into a
database. Then if the same person has committed a crime then the officials can run a
diagnostic test to find the culprit.
 ➢ Implementation of this program in public vehicles or private taxis is very much useful
since it can identify the face structure and store it in a database, if the face structure is
already been saved in the database then it runs a background check to see if there is any
prevailing crimes listed on the individuals name. This can be then notified to the driver as
well as the near by police station which improves the safety of the driver.
➢ The alert app technology invented by Leeway can be also integrated with the program to
get a immersive look into the alert system.

DISCUSSION & SOLUTIONS

The integration of alert app technology invented by Leeway can be a immense game
changer as it provides in the best mode of alerting method. The addition of face structure to the
database can also be help for the police department to identify the criminal or the person who is
guilty. This service of the program can be used for the surveillance of the city or the
neighborhood so that there are mainly two benefits for this as firstly they can identify who is not
wearing mask and secondly if they are wearing masks and then committing a crime they can be
caught quite quickly.
CONCLUSIONS
Implementation of the above mentioned face detector can reduce the crime rates tremendously so
its all the way the best solution for reduction of petty crimes and for surveillance of the city or
the place the program has been implemented.
REFERENCES
Journal:
1. Arijit Malakar; Ashok Kumar; Sudipta Majumdar, ‘Detection of Face Mask in Real-time using
Convolutional Neural Networks and Open-CV’, Published in: 2021 2nd International Conference
for Emerging Technology (INCET).
2. Huaizu Jiang and Erik Learned-Miller, ‘Face Detection with the Faster R-CNN’, 12th IEEE
International Conference on Automatic Face & Gesture Recognition, 2017.
3. Saad Albawi, Tareq Abed Mohammed and Saad Al-Zawi, ‘Understanding of a convolutional
neural network’, IEEE International Conference on Engineering and Technology (ICET), 2017.
4. Matthias Dantone; Juergen Gall; Gabriele Fanelli; Luc Van Gool, ‘Real-time facial feature
detection using conditional regression forests’, Published in: 2012 IEEE Conference on
Computer Vision and Pattern Recognition Referenced in:IEEE Biometrics CompendiumIEEE
RFIC Virtual JournalIEEE RFID Virtual Journal.
5. Wang Chen S, Horby Peter W, Hayden Frederick G, Gao George F (2020) A novel
coronavirus epidemic of global concern for health. It’s the Lancet. 395(10223):470–473.
6. S. Yadav, "Deep learning based safe social distancing and facemask detection in public areas
for COVID-19 safety guidelines adherence", Int J Res Appl Sci Eng Technol., vol. 8, no. 7, pp.
1368-1375, 2020.
7. M Loey, G Manogaran, MHN Taha and NEM. Khalifa, "A hybrid deep transfer learning
model with machine learning methods for face mask detection in the era of the COVID-19
pandemic", Measurement (Lond), vol. 167, pp. 108288, 2021.
8. S Akshay, Mandara Bhat and Aishwarya Rao, "Facial Expression Recognition using
Compressed Images", International Journal of Recent Technology and Engineering, 2019.