Authors :
M. Murali; Deepak Yadav
Volume/Issue :
Volume 7 - 2022, Issue 4 - April
Google Scholar :
https://bit.ly/3IIfn9N
Scribd :
https://bit.ly/36SSN19
DOI :
https://doi.org/10.5281/zenodo.6499706
Abstract :
COVID-19 is a rapidly spreading viral disease
which has challenged the health services of the world.
Social distancing has been recommended as one of the
best practice that helps to restrain the curve of COVID19 virus. The effective measure of Social distancing has
helped to decrease the transmission rate of the infectious
COVID-19 worldwide.Furthermore, the lack of
temporalunderstanding among the people may cause
unintentional breach of the social distancing norms.
Hence, it is necessary to bring in a vision based
concurrent flow that will spot the social distancing
violations. Social Distancing limits the physical contact
among the people and by doing such, the danger of
spreading COVID-19 can be decreased. The main
objective of this proposed systemis to create a deeplearning system to detect social distancing to recognize
persons in video sequences. The proposed system will
employ YOLOv3 object recognition algorithm. The
significance of this model is improvised through the
transfer learning process. The pre trained algorithm is
coupled with the trained layer which uses an additional
data that will help in the detection process. The
Euclidean distance is used to compute the pairwise
distances of objects from the identified bounding box
centroid while the bounding box information helps to
identify the objects.A social distancing violation
threshold will beset to examinewhether the distance
value among the people exceeds minimal barrier that has
been set for social distance. This work will define a social
density value and show that pedestrian-density is
heldunder the value defined. Thus thechance of a Social
Distancing violation could be prevented.
COVID-19 is a rapidly spreading viral disease
which has challenged the health services of the world.
Social distancing has been recommended as one of the
best practice that helps to restrain the curve of COVID19 virus. The effective measure of Social distancing has
helped to decrease the transmission rate of the infectious
COVID-19 worldwide.Furthermore, the lack of
temporalunderstanding among the people may cause
unintentional breach of the social distancing norms.
Hence, it is necessary to bring in a vision based
concurrent flow that will spot the social distancing
violations. Social Distancing limits the physical contact
among the people and by doing such, the danger of
spreading COVID-19 can be decreased. The main
objective of this proposed systemis to create a deeplearning system to detect social distancing to recognize
persons in video sequences. The proposed system will
employ YOLOv3 object recognition algorithm. The
significance of this model is improvised through the
transfer learning process. The pre trained algorithm is
coupled with the trained layer which uses an additional
data that will help in the detection process. The
Euclidean distance is used to compute the pairwise
distances of objects from the identified bounding box
centroid while the bounding box information helps to
identify the objects.A social distancing violation
threshold will beset to examinewhether the distance
value among the people exceeds minimal barrier that has
been set for social distance. This work will define a social
density value and show that pedestrian-density is
heldunder the value defined. Thus thechance of a Social
Distancing violation could be prevented.
