Scribd Logo
Upload

Welcome to Scribd!

  • 51 views

    Social Distancing Detector Using Yolo

    Uploaded by

    Pavan kumar Kotapally
    This document summarizes a project to develop a deep learning model using YOLO (You Only Look Once) for detecting social distancing violations. The model uses YOLO for object detection to identify people in video frames and place bounding boxes around each person. It then performs non-maxima suppression to reduce overlapping boxes. The centroids of each box are used to calculate pairwise distances between people and check if distances are below a threshold, indicating a violation of social distancing guidelines. The goal is to help enforce social distancing in public areas and minimize the spread of COVID-19.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd

    Social Distancing Detector Using Yolo

    Uploaded by

    Pavan kumar Kotapally
    51 views4 pages
    This document summarizes a project to develop a deep learning model using YOLO (You Only Look Once) for detecting social distancing violations. The model uses YOLO for object detection to identify people in video frames and place bounding boxes around each person. It then performs non-maxima suppression to reduce overlapping boxes. The centroids of each box are used to calculate pairwise distances between people and check if distances are below a threshold, indicating a violation of social distancing guidelines. The goal is to help enforce social distancing in public areas and minimize the spread of COVID-19.

    Original Title

    somepaper

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document summarizes a project to develop a deep learning model using YOLO (You Only Look Once) for detecting social distancing violations. The model uses YOLO for object detection to identify people in video frames and place bounding boxes around each person. It then performs non-maxima suppression to reduce overlapping boxes. The centroids of each box are used to calculate pairwise distances between people and check if distances are below a threshold, indicating a violation of social distancing guidelines. The goal is to help enforce social distancing in public areas and minimize the spread of COVID-19.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    51 views4 pages

    Social Distancing Detector Using Yolo

    Uploaded by

    Pavan kumar Kotapally
    This document summarizes a project to develop a deep learning model using YOLO (You Only Look Once) for detecting social distancing violations. The model uses YOLO for object detection to identify people in video frames and place bounding boxes around each person. It then performs non-maxima suppression to reduce overlapping boxes. The centroids of each box are used to calculate pairwise distances between people and check if distances are below a threshold, indicating a violation of social distancing guidelines. The goal is to help enforce social distancing in public areas and minimize the spread of COVID-19.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 4

    SOCIAL DISTANCING DETECTOR USING YOLO

    Balineni Kushaal
    [email protected] Naveen Marella Penugonda Sandeep
    [email protected] [email protected]

    halls, shopping malls, airports, schools,etc).social


    ABSTRACT— So as the corona virus(covid-19) distancing mainly benefits the people those are at higher
    pandemic has brought the catastrophe with its rapid risks with pre medical conditions.
    spreading. In the present scenario as there are no
    Our main objective through this project we aim to develop a
    effective proper medical treatments aganist the virus, it
    deep learning platform for detection of people. In order to
    may cause threat to humanity .Therefore, social distancing detect social distance at controlled environments. As we
    is the prominent way to stop the increase in positive need public to follow social distancing we need a system to
    cases of covid-19. And the virus risk can be brought under observe and find those who violate. . So by minimizing the
    control by maintaining physical distance from each spread from one person to person, disease spread can be
    individual. The aim of this project is to improve a deep minimized.
    learning technology for social distancing algorithm using What is YOLO?
    YOLO . Due to its good accuracy towards the detection of The abbreviation of YOLO is given as YOU ONLY LOOK
    subjects (humans,objects) and small objects, in this ONCE . It is a real time object recognition algorithm, first
    project we have used yolov3 . Due to its adjusted network mentioned and written in the 2015 seminar paper by Joseph
    structure Yolo v3 has an advantage that is used for multi Redmonetal yolo involves the concept of object detection
    scale object detection . our dectecting model identifies the with higher accuracy rate.
    people placing bounding boxes over the subjects . This
    methodology of using yolo increases the accuracy of the
    model when compared to RCNN&SSD as shown in below
    fig.

    INTRODUCTION I. LITERATURE SURVEY


    . This project was espoused and developed due to two- In the proposed model [3] the authors explained and
    pronged reasons : developed the Tinier YOLO, which is derived from the
    1. An intelligent model to identify the persons YOLO-V3, it is a model suggested for obtaining better
    without mask. accuracy of detection by minimizing the model size and
    2. A socially impactful model to stop the spread of real-time performance. From this journal we have obtained
    virus. the idea and how to use the COCO data set.
    The  present COVID pandemic, also known as
    the coronavirus is still a viral disease which we are From [1] authors used the single stage network framework
    facing since 2019 and to as of date , it is caused of an to estimate the overlapping boxes and their class
    acute respiratorysyndrome SARS COV 2.The virus was probabilities.
    first discovered in December 2019 in wuhan(china). The At first COCO was used to train the model . Full form of
    WHO declared the outbreak as a global pandemic on COCO is “Common objects in context”.
    11 March.As of April 30 2021, more than 150  million COCO is one of the database management system. In this
    cases  of COVID-19 have been reported resulting paper we analyzed the concept and approach in develop of
    in more than 3.17 million deaths . the code.
    Social distance will help maintain distance between each
    individual in order to cut down the increase of covid19 The NMS (NON-MAXIMA SUPRESSION) algorithm [5] is
    spread it helps in various crowded places like (function explained how to use. Which is used to decrease the
    overlapping of the multiple bounding boxes. So that the box
    which contains the more probability can be drawn around assigned in video frame as shown in the below fig (a) . In
    the subject this project We used YOLO V3 .

    The estimation and characterization of person-person


    distances from images is presented as the VSD problem
    Solving[7] such issues enables a rapid monitoring of the
    population for possible threats .behaviour that poses a health
    risk, particularly in light of recent events Outbreaks of
    pandemics

    From [4] The output of commonly used deep learning


    techniques for object detection and recognition are in pattern
    of RGB(RED-GREEN-BLUE). So we have also used the
    same RGB pattern. Thermal surveillance scenarios were
    used to analyse visual images.
    Fig(a): Schematic representation of YOLO v3 architecture
    II. PROPOSED WORK
    Non-Maxima suppression
    Initially it detects the subjects(“humans”and etc) in a
    video frames by using YOLO object detection algorithm. NMS (Non-maxima suppression) is a technique used in
    The primary goal of this device is to analyse recorded video many computer vision algorithms also used to reduce
    footage for individual identification and then process it overlapping bounding boxes to only a single bounding box,
    further for social distancing or safety violations. It is used to thus representing the true detection of the object. Having
    give the output of the person prediction’s probability, overlapping boxes for each individual is not accurate
    boundary box coordinates for the detected person, and the especially when we are counting the number of boxes in
    centroid of each individual. each individual frame. The example of NMS is given in the
    This process contains 4 stages below fig b where the algorithm detects the bounding box
    1. Object detection around the object which has maximum probability by
    suppressing the minimum probabilities(i.e which are not not
    2. Non-Maxima suppression. maximum). So from the example the box with 0.9
    3. Object localization probability is considered as the bounding box.
    4. Distance calculation.

    III. METHODOLOGY

    Object detection

    YOLO OBJECT DETECTOR

    With just one look at the picture, the type of an object and
    location of the object can be predicted by using this model.
    In allocating class probabilities to the subjects, YOLO treats
    the detection issue as a regression task rather than a
    classification task. This YOLO algorithm also predicts
    multiple bounding boxes for a single subject and with
    closest probabilities at the same instance. Basically
    YOLOv1, YOLOv2, and YOLOv3 are the three main Fig (b) showing an example of Non maxima suppression
    algorithms used in object classification. The YOLO v1
    algorithm is used to identify subjects in the video frame
    image. YOLO is an object detector which can detect the Distance caluculation
    possible areas within an image where certain objects may be
    present and classifies certain areas into object classes in a After detection of each individual in each video frame, in
    single step. YOLOv3 has a 53 convolutional layers further step, the distance is computed by the bounding boxes
    including 3x3 and 1x1 filters, with shortened connections from the centroid of every identified subject. The detected
    between each layers for extraction of each subject. For each boundary box specific coordinates are used for calculating
    individual subject three predictions of bounding boxes are each box’s centroid. After calculating, the centroid each
    identified bounding box Is given a unique ID and the next
    step is to compute the distance between each detected
    object. we first examined each individual frame in the video
    stream and centroids of every bounding box and the pair-
    wise distance is calculated using Euclidean distance. Every
    centroid data is stored within the sort of an inventory. Based
    on the each distance value between every pair in the frame To calculate the distance between the centroids
    image an input threshold value is given to check if the
    distance is less than the specified pixel or not. If they are
    C1 and C2 of the detected pair subjects
    less than given threshold value the colour of the bounding
    box is displayed as red indicating they are not following the
    social distancing or else it displays green as shown in the
    flow chart diagram in fig(c) The information in the violation
    set will be updated frame to frame. If the violation set (list)
    consists the existing present index , then the colour is
    where c= centroid of the bounding box,
    changed to red, and even the number of violations is also C1=(Xmin,Ymin) ; C2=(Xmax,Ymax) , (x,y) are
    shown in the output. the co-ordinates of the centroid.
    IV. FLOW CHART

    VI. RESULTS DISCUSSION

    In Fig1 and Fig2 , the testing results of the social distancing


    detector are given below . The results obtained are updated
    at every individual frame , providing us with the accurate
    results for spreading of virus among the people.

    Fig 1

    Fig(c)
    Fig 2

    V. EQUATIONS & PARAMETERS VII. CONCLUSION

    As Artificial intelligence technology is blooming in today


    To calculate the bounding box center point C(x,y) world, we have social detector which can contribute to the
    the formula for midpoint of the detected person in healthcare of the world. This model consists of YOLO as the
    backbone which can detect the HUMANS. With this project
    the frame is given as we have developed a social distancing detector using deep
    learning concept. YOLOv3 trained with coco dataset is used
    for human detection to obtain good accuracy for detection
    .The project can be developed in the future for detecting in Sell Decision’’ Model Over 2D Candlestick Charts"
    all type of environments where social distance is must to be Published 2020 may 14.
    maintained. By the development of Social distancing
    detector we can detect if the person is in close proximity 6. Afiq Harith Ahamad, Norliza Zaini, Mohd Fuad Abdul
    and it would be of great help to the society. Latip "Person Detection for Social Distancing and Safety
    Our future work will extend the object localization Violation Alert based on Segmented ROI"2020 10th IEEE
    algorithm with further more features and testing the InternationalConference. 2018;5:1598– 1605.
    algorithm with high performance GPU’s for better video
    outputs. 7. MARCO CRISTANI,ALESSIO DEL BUE,VITTORIO
    MURINO,FRANCESCO SETTI "The Visual Social
    Distancing Problem" Published 2020.

    VIII. REFERENCES 8. Ahmed I., Ahmad A., Piccialli F., Sangaiah A.K., Jeon G.
    IEEE Internet of Things Journal. 2018;5:1598– 1605.

    1. Imran Ahmed, Misbah Ahmed ,joel J.P.C. Rodrigues, 9. Ahmed I., Ahmad M., Adnan A., Ahmad A., Khan M.
    Gwanggil Jeon, and Sadia Din ‘A deep learning based social International Journal of Machine Learning and Cybernetics.
    distancing monitoring framework for COVID-19’ 2020 Nov 2019:1–12.
    1. pp.7-20
    10. Girshick R. Proceedings of the IEEE international
    2. Ahmad M., Ahmed I., Ullah K., Khan I., Adnan A. 2018 conference on computer vision. 2015. pp. 1440–1448
    9th IEEE annual ubiquitous computing, electronics mobile
    communication conference (UEMCON) 2018. pp. 746–752 11. Ramadass L., Arunachalam S., Sagayasree Z.
    . International Journal of Pervasive Computing and
    3. WEI FANG,LIN WANG,and PEIMING REN " Tinier- Communications. 2020
    YOLO: A Real-Time Object Detection Method for
    Constrained Environments" Published on 2019 Dec 24. 12. N. S. Punn, S. K. Sonbhadra, and S. Agarwal, “Covid-
    19 epidemic analysis using machine learning and deep
    4. MATE KRIŠTO,MARINA IVASIC-KOS,and MIRAN learning algorithms,” medRxiv, 2020.
    POBAR "Thermal Object Detection in Difficult Weather
    Conditions Using YOLO" 2020 july 06 vol 8.pp.8-9

    5. SERDAR BIROGUL,GÜNAYTEMÜR,and UTKU


    KOSE "YOLO Object Recognition Algorithm and ‘‘Buy-

    You might also like