SOCIAL DISTANCING DETECTION

INTERNATIONAL RESEARCH JOURNAL OF ENGINEERING AND TECHNOLOGY (IRJET) E-ISSN: 2395-0056
VOLUME: 09 ISSUE: 03 | MAR 2022 WWW.IRJET.NET P-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1305
SOCIAL DISTANCING DETECTION
Shivani Gupta1, Shivani Kale2, Ruchita Keni3, Prof. Deepali Maste4
1Shivani Gupta, Dept of Information Technology Engineering, Atharva College of Engineering
2Shivani Kale, Dept of Information Technology Engineering, Atharva College of Engineering
3Ruchita Keni, Dept of Information Technology Engineering, Atharva College of Engineering
4Prof. Deepali Maste, Dept of Information Technology Engineering, Atharva College of Engineering
Maharashtra, India
----------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - In the world of clever technology the place where
everything is being built using technology. As technology is
enhancing day by day the need for technical systems is
additionally growing.
The paper presented here is a user-friendly deep learning
social distancing detection that simplifies the task of
preserving safe distance between each other. The foremost
intention of this project was to alert pedestrians to maintain
a safe distance. The video frame or a photo is given as input.
The real-time object detection YOLO algorithm which was
pre-trained using the COCO dataset detects the objects I.e.
pedestrians from the frame by the usage of bounding box
regression. Moreover, using the centroid points of the
bounding boxes the distance between them is being
estimated using the centroid formula. The pedestrians
violating the default distance are displayed in red bounding
boxes and the ones who are no longer violating are displayed
in green. Thereafter, it displays the whole no. of violations
and alert messages on the screen. The technique was once
established on a pre-recorded video of pedestrians as well as
an image.
The result suggests that the technique is in a position to
determine no. of violations in the frame and additionally
helps to maintain social distance.
Keywords-YOLO algorithm,COCO dataset, Social
distancing detection,Bounding box regression
I.INTRODUCTION
As we all are conscious that COVID-19 instances are rolling
out throughout the world day by day. The COVID-19 virus
can unfold from a contaminated person's mouth or nostril
in small particles when they cough, sneeze, speak, sing or
even breathe. At this, the term social distancing appears to
be spoken more, be it social media or any news channel.
This idiom appears to be immersing all over. And the first-
rate way to sluggish down the transmission is to inculcate
the social distancing manner in our life. Also, as suggested
by WHO, to shield ourselves and others through
contamination we need to remain at least 6 feet away from
each other. But it appears to be aggravating to keep
impervious distance in this pandemic.People would
possibly be keeping distances and some would perchance
not. So, to make it easier, we have designed a social
distancing detection system. It is a real-time object
detection computing device that simplifies the work of
preserving social distance.
This device is designed to aid the users to retain a secure
distance from each other. It helps the users by exhibiting
the no. of violations and alert messages. Hence, users will
be capable of holding the social distance.
II.PROPOSED APPROACH
As we all know, the imitable condition that is going on due
to the COVID-19 pandemic. And in this pandemic, retaining
a protected distance manually is such a disturbing task.
People want to be maintaining distances, and some would
perchance not. So, to make it easier, we have designed a
social distancing detection system that helps users to keep
their social distance from others. We have used the YOLO
algorithm which is pre-trained by the usage of the COCO
dataset. YOLO algorithm is a real-time object detection
algorithm that determines the bounding boxes and
classification of the objects precisely. YOLO firsts divide
the image into N grids. Each N grid has an equal dimension
location of S*S. After this, image localization and detection
is utilized on the grid. The distance between the centroid
elements is estimated with the use of the centroid formula.
Thereafter, distance is in distinction to the default distance
and the pedestrians who are close to each other are
displayed into red bounding boxes and the ones who are
no longer are displayed into the green. Therefore, the
whole no. of violations and an alert message is displayed
on the screen. By displaying the alert message, users will
be able to maintain a safe distance from each other.

III.ARCHITECTURE
Fig-1- System Architecture
The user needs to give the digital camera frame or a pre-
recorded video and image can be given as input. After that,
the real-time object detection mannequin YOLO algorithm
pre-trained using the COCO dataset will examine the
pedestrians from the frame through the usage of bounding
box regression. Thereafter, the system will estimate the
distance between them by the usage of the centroid
elements of the bounding box. The pedestrians who are
close to each other are displayed in red bounding boxes
and the ones who do not are displayed in green. As a result
the whole no. of violations and alert messages is displayed
on the show display screen.
IV. METHODOLOGY
1. YOLO algorithm: YOLO is generally a real-time object
detection algorithm that follows a regression approach
which essentially is quite significant. This algorithm helps
to discover the bounding boxes as nicely as the class of the
object subtly specifying its location. YOLO algorithm first
splits the frame into N framework where each framework
is having equal dimension region of S*S. These N
frameworks are liable for image localizing and
determining. Moreover, the objects are determined from
the frame which is quite significant. Because of its speed
and accuracy, the objects are determined in only one
execution of the YOLO algorithm in a sort of major way.
Hence, this is the reason the YOLO algorithm outperforms
all other models.
2. COCO Dataset: Common objects in Context or COCO is
particularly a large-scale object detection segmentation
and a captioning dataset which for all intents and purposes
is fairly significant. It consists of a couple of sets of
excessive datasets each made for a unique machine
learning task. Algorithms for object detection and
classification can essentially be pre-trained using the
dataset in a major way. The COCO dataset has a specific
layout that exactly defines how your notations i.e.
bounding boxes, objects, classes, etc. are saved on disk. The
COCO dataset is deliberately biased towards person class,
so it determines person class more accurately which is
quite significant.
3. Python: Python is a markup language, which is
specifically practicable. It is advised to be mostly applied in
a fluctuation of functions such as big data, software
program, internet improvement, computerization and
normally getting tasks executed for the most part. It is used
to automate tasks to construct websites, creating a variety
of programs. Python helps programmers to complete
complex tasks without encountering coding problems
which is significant.
4. NumPy: NumPy may be a multipurpose library of python
which is employed for handling arrays. NumPy essentially
stands for Numerical Python which functions in the
domain of algebra. It gives array objects quicker than the
python list. Since; NumPy may be a library of python but
it's partially written in Python and for computations, C
language is employed.
5. SciPy: SciPy, which specifically symbolizes scientific
python, is particularly an open-source public library of
python that makes use of NumPy underneath. It mostly has
additional features that are often to be used for very large
statistics and computer learning. SciPy can also be used for
performing complicated numerical operations,
optimization, integration, graphic processing, etc.
6. Imutils: Imutils mostly is a sequence of features that
helps to make primary image processing functions such as
resizing, translation, relocation, rotation, extraction of the
skeleton, picture scaling, etc. It is a python package, which
particularly is in all likelihood to be based on. It is able to
call the OpenCV interface simply.
7. OpenCV: OpenCV is an open-source library that is used
for picture processing, computing device learning, and
computer vision. OpenCV has additionally been used in
true time operations, the place where it makes the
responsibilities greater and easier. Using OpenCV we can
manage films and snapshots from which the objects,
human faces, and even handwriting can be identified.
When OpenCV is built-in with one of a kind libraries it
affords
V. IMPLEMENTATION
The system captures the image then using the YOLO
algorithm and COCO dataset it will detect the people from
the frame. Thereafter using the centroid formula the
distance between all detected people will be calculated.
Based on these distances, if two people are less than
provided distance system will generate red boxes around
them and green for others, System will show the number of
social distancing violations performed in the frame on the
screen and also if the violations are greater than the

specified number then alert message will be displayed on
the screen.
VI. ADVANTAGES
1.Conventional video supervision computing system
requires security officers to display the video screen,
which can moreover lead to false results which isn’t
trustworthy and there are possibilities that some would
possibly violate the distance so to avoid this we have
designed a system that does not require human beings to
monitor the screen, it captures the video itself and
recognize the pedestrians that are violating which is now
no longer possible manually.
2. We have designed a social distancing detection system
that simplifies the work of retaining impenetrable social
distance.
3. We have used the YOLO algorithm which is a real-time
object detection algorithm that detects the bounding boxes
and class of the image which consists of objects internal in
only one execution of the algorithm.
4. COCO dataset, which is an open-source high first-rate
dataset. It is immoderately biased towards person class, so
it detects its person class precisely
5. Social distancing detection system shows no. of
violations and alerts users via displaying the alert message.
6. YOLO algorithm works properly due to the actuality that
it is pre-trained with the usage of the COCO dataset.
7. YOLO algorithm which we have used outperforms all
other models due to the actuality of its pace and accuracy.
VII. RESULT
The YOLO algorithm works properly in real-time object
detection operations. It determines the bounding boxes
and class of the image that consists of objects in solely one
execution of the algorithm. It gives a lot of fairly higher
performance in a generally big way. Also, the system will
help users by displaying the no. of violations and alert
messages, so that they can keep a secure distance. It also
displays the pedestrians that are violating red bounding
boxes and who are not into the green. All this is proven in
the images illustrated below.
Fig-2: Output-1
Fig-3: Output-2
Fig-3: Output-3
VIII. CONCLUSION
The system was designed keeping in mind the simplicity of
use via the users so that users will particularly be able to
use the device in their everyday life in a subtle way. Social
distancing detection is an environment-friendly real-time
object detection device that essentially helps users in
maintaining a safe distance. The foremost aim was to help
users in retaining a safe distance and alert the ones who
are violating by showing the alert message on the screen.
This system proposes a very high-quality deep gaining
knowledge of the system that simplifies the work of
preserving social distance using the YOLO algorithm which

was pre-trained through the COCO dataset. The result
indicates that the proposed approach is supposed to be
used in any working surroundings because of its velocity
and accuracy
REFERENCES
1. E. Rehder, F. Wirth, M. Lauer, and C. Stiller, “Pedestrian
prediction by planning using deep neural networks,” in
2018 IEEE International Conference on Robotics and
Automation (ICRA), pp. 1–5, IEEE, 2018.
2. R. Q. M´ınguez, I. P. Alonso, D. Ferna´ndez-Llorca, and
M. A´ . Sotelo, “Pedestrian path, pose, and intention
prediction through gaussian process dynamical
models and pedestrian activity recognition,” IEEE
Transactions on Intelligent Transportation Systems,
vol. 20, no. 5, pp. 1803–1814, 2018
3. B. Galdino and A. Nicolau, “A measure distance system
for docks: An image-processing approach,” in 2017
IEEE First Summer School on Smart Cities (S3C), pp.
145–148, IEEE, 2017
4. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A.
(2016). You only look once: Unified, real-time object
detection. In Proceedings of the IEEE conference on
computer vision and pattern recognition (pp. 779-
788).
5. R. Q. M´ınguez, I. P. Alonso,D. Ferna´ndez-Llorca, and.
Sotelo, “Pedestrian path, pose, and intention
prediction through gaussian process dynamical
models and pedestrian activity recognition,” IEEE
Transactions on Intelligent Transportation Systems,
vol. 20, no. 5, pp. 1803–1814, 2018.
6. N. J. Singh and K. Nongmeikapam, “Stereo system
based distance calculation of an object in image,” in
2019 Fifth International Conference on Image
Information Processing (ICIIP), pp. 29–34, IEEE, 2019.
7. C. H. Lam and J. She, “Distance estimation on moving
object using ble beacon,” in 2019 International
Conference on Wireless and Mobile Computing,
Networking and Communications (WiMob), pp. 1–6,
IEEE, 2019.
8. B. Galdino and A. Nicolau, “A measure distance system
for docks: An image-processing approach,” in 2017
IEEE First Summer School on Smart Cities (S3C), pp.
145–148, IEEE, 2017.
9. MChehab, Android – Calculating Distance Between
Two Points, 2009 (accessed February 3, 2014).
10. Y. You and C. Wu, “Indoor positioning system with
cellular network assistance based on received signal
strength indication of beacon,” IEEE Access, 2019

SOCIAL DISTANCING DETECTION

More Related Content

What's hot (20)

Similar to SOCIAL DISTANCING DETECTION (20)

More from IRJET Journal (20)

Recently uploaded (20)

SOCIAL DISTANCING DETECTION