Abstract:
Covid-19 dissemination can be slowed via social distancing strategies. Social distance
is closely followed as a model to break the chain of dissemination. This study offers a
method for detecting social distance violations in public settings such as shopping
centers, marketplaces, Banks, and clinics. It would be simple to monitor individuals
using this model to see if they are maintaining social distance and to notify them if any
deviations from the predefined boundaries occur. From an overhead perspective, the
purpose of this project is to construct a deep learning platform for forecasting social
distance after victimization. The system uses the YOLO v3 object recognition model to
detect humans in video or picture sequences. The transfer learning technique is also
utilized to increase the model's accuracy. The detection algorithmic rule in this method
makes use of a well before algorithmic rule that's linked to a second trained layer that
exploits an overhead human knowledge set. The bounding box information is used by
the detection algorithm to identify who has been victimized. The bounding box
information is used by the detection model to identify humans. The purpose of this
project is to develop a victimization deep learning platform. This algorithm can also be
used to perform the procedure on real CCTV footage. The model employs deep learning
techniques with the distance in between individuals in the frame and the YOLO v3
model trained just on COCO dataset to identify individuals in the video. The system is
designed to fit the environment in which it will be used. YOLO model, COCO dataset,
Image processing, Deep learning, and so on are examples of index terms. The distance
between the individuals is determined using the associate degree object detection
algorithmic rule. A Euclidian distance is calculated and compared to the quality
distance provided. A green-colored bounding box will represent those with enough
distance, while a red-colored bounding box will indicate those with inadequate distance.
The complete range of breaches is also displayed in the output stream. This study
intends to halt the deadly virus's spread to a considerable extent with minimal human
effort and a low chance of disease for local governments.
