SlideShare a Scribd company logo
1 of 25
By Aparnaa Senthilnathan, High School
Senior, Bronx Science
AreYouSocial
Distancing?A
Computational
Approach
“We’ve actually invested very little in a
system to stop an epidemic...The problem
wasn’t that there was a system that didn’t
work well enough. The problem was that we
didn’t have a system at all”
—BillGates
”
Introduction
01.
Research Goals and Hypothesis
Research Goals
Successfully apply mask
recognition to videos
Use ML algorithms to influence effective
measures to stopping the pandemic
Use as an analytical tool for calculating
the percentage of masks
Implement the program as a
security measure in buildings
If CNN models are trained on a few mask and no mask
datasets and then tested on separate images and
videos… then:
• It will correctly identify masks and an alarm will
sound that will remind non-mask wearers.
• This program will correctly predict whether people
are six feet or more apart.
• This method will be more effective in raising
awareness towards pandemics and will be ready
for scientists and researchers to use the program
as both an analytical tool and a security measure
in buildings.
Hypothesis
Methods
02.
CNN Models, First Times?, Equations, Main Menu Module, Details
● Neural Networks
● Used to train the CNN model to
identify items in images
● Consists of four layers: convolutional,
pooling, fully connected and softmax.
● These layers are ordered by how they
learn filters of increasing complexity.
CNNModels
• Use machine learning concepts to create a program to help encourage more people
to follow the guidelines necessary for Coronavirus prevention
• Attempt to implement mask recognition algorithms into video recordings.
Firsttimes?
Materials
The program was coded using
Python and its libraries. A few
mask and no mask datasets
from Kaggle and Github were
download.
TSC is the number of test
samples the model was able to
correctly identify and the TS is
the total number of test samples
received by the model.
The loss is just calculated by
subtracting the accuracy from 1,
where 1 is the maximum
percentage of accuracy
Accuracy Loss
Equations
A(TSC,TS)
= TSC/TS
L(A) = 1 - A
P is the percentage of people
wearing masks, M is the number
of masks identified, and NP is the
number of people in the frame.
Percentageofmasks
Equations(cont.)
P = M/NP
TheProgram’s mainMenuModule
Take a Picture
using the
Computer
Open a
Webcam Video
on your
Computer
Upload an
Image
Upload a Video
01 02 03 04
Exit
05
Demo
The photo sent by the user
Percentage of mask-
wearers over a total
number of people
A green border indicates
the person is wearing a
mask
If the person isn’t wearing
the mask, then the border
should be red.
Results
Observations and Data Collection
03.
Table1:TakeaPhoto
Trials(MaskPercentage/#ofPeoplerecorded/TotalFacesshown)
Photo Type Trial 1 Trial 2 Trial 3
Single Person (Mask) 0%/1(plastic mask) 100%/1 (pink
mask)
100%/1 (black
mask)
Single Person (No-Mask) 0%/1 0%/1 0%/1
Two People (Both Masks) 50%/2 100%/2 50%/2 (-wrong
position)
Two People (One Mask) 50%/2 (wrong face) 100%/2 (person
without a mask
counted)
50%/2
Two People (No Masks) 0%/2/2 0%/2/2 0%/2/2
Group of People (All Masks) 66.7%/2/3 100%/3/3 33.3%/1/3
Group of People (Some Masks) 66.7%/3/3 66.7%/3/3 33.3%/1/3
Group of People (No Masks) 33.3%/3/3 (one non-mask wearer counted) 33.3%/3/3 (one
non-mask wearer
counted)
33.3%/3/3 (one
non-mask wearer
counted)
Table2:UploadaPhoto
Trials(MaskPercentage/DistanceAccuracy/#ofPeoplerecorded/Total
Facesshown)
Photo Type Trial 1 Trial 2 Trial 3
Single Person (Mask) 100%/100%/2 (pillow counted/1) 100%/100%/1/1 100%/100%/1/1
Single Person (No-Mask) 0%/100%/1/2 0%/100%/1/2 0%/100%/1/2
Two People (Both Masks) 100%/100%/2/2 100%/100%/2/2 100%/100%/2/2
Two People (One Mask) 100%/100%/1/2 0%/0%/0/2 50%/100%/2/2
Two People (No Masks) 100%/100%/2/2 50%/100%/2/2 50%/100%/2/2
Group of People (All Masks) 0%/100%/2/4 75%/50%/8/13
(girl’s shoulder
counted)
86.66%/100%/15
/50
Group of People (Some Masks) 100%/0%/5/20 (a non-mask wearer counted) 53.85%/100%/13
/13 (phone
counted)
70%/100%/28/30
(few non-mask
wearers counted)
Group of People (No Masks) 0%/0%/10/10 0%/0%/9/9 0%/0%/6/7
Table3:Webcam
Trials(MaskPercentage/PeopleIdentified/Numberof
People)
Photo Type Trial 1
Single Person (Mask) 0%/1/1
Single Person (No-Mask) 100%/1/1
Two People (Both Masks) 50%/2/2
Two People (One Mask) 50%/2/2
Two People (No Masks) 0%/2/2
Group of People (All Masks) 33.3%/1/3
Group of People (Some Masks) 66.7%/2/3
Group of People (No Masks) 0%/3/3
Table4:Upload aVideo
VideoName Observations
How to wear your single use face mask? (short version) Before the girl puts on her mask, the program identifies that
her face has a mask on. As the girl was putting on her mask,
the camera turned 360 degrees in slow motion, which the
program wasn’t able to identify her until her face was shown
in front view. The program was able to identify her with the
mask on correctly, but without the mask, the program still
identifies her as her wearing the mask.
Group of People Putting on Masks Three people are recorded in front view by the camera, and the
program was able to classify which of the three people are
wearing masks as well as the distance between them for the
most part.
dISCUSSION
What can we infer?/Implications
04.
Observations
Photos
● There were mixed results
● The program was able to identify and classify masks correctly for the most part.
● The program also calculated the percentage of masks within the image correctly.
● In some cases, the program can’t identify other faces due to lighting conditions or the person is wearing
the mask
Videos
● There also were mixed results
● The program can identify and classify the face properly when the subject is facing the camera.
Otherwise, it is hard for the program to identify the face.
● The program also calculated the percentage of masks within the video correctly.
● The alarm is correctly initiated when the program identified a non-mask wearer in the webcam video.
Mask recognition can be used
successfully in videos to an
extent
For the most part, the program
can serve as an analytical tool
because it calculates the
percentage of masks correctly.
VideoRecognition AnalyticalTool
Discoveries
The program allowed the pre-
trained model to identify an
object as a face, despite a low
confidence score.
If the video size is large, the
video playback would slow down
a lot due to the excessively large
amount of data the model has to
read.
Wrongface Videoprocessing
Limitations
No mask
• Find ways to improve the machine learning
algorithms such that the program can process
videos in original sizes and allow the program to
identify the right faces in the same video.
• Can be expanded further by training the model to
identify whether or not the person is wearing the
mask correctly and is it the right type of mask.
• Introduce future methodology that allow
scientists to jump further into raising awareness
towards the coronavirus pandemic.
Future RESEarch
• All icons and images are created by Flaticon and Freepick. The screenshots of
code/demo are from my laptop. The presentation template was created by
Slidesgo.
• Special thanks to Dr. Sarah Ostadabbas of Northeastern University and members
from the MIT Beaver Works Summer Institute for mentoring and helping me with
CNN Networks and resources necessary for the project.
• Also, a big thanks to Dr. Vladimir Shapovalov for guiding me through the research
process
Acknowledgements
1. The University of Texas at Austin. "Coronavirus spreads quickly and sometimes before people have symptoms, study finds." ScienceDaily. ScienceDaily, 16 March 2020.
<www.sciencedaily.com/releases/2020/03/200316143313.htm>.
2. Bai, Nina. “How the New Coronavirus Spreads and Progresses – And Why One Test May Not Be Enough.” How the New Coronavirus Spreads and Progresses – And Why One Test May Not Be
Enough | UC San Francisco, 10 Nov. 2020, www.ucsf.edu/news/2020/02/416671/how-new-coronavirus-spreads-and-progresses-and-why-one-test-may-not-be-enough.
3. The New York Times. “Covid World Map: Tracking the Global Outbreak.” The New York Times, The New York Times, 28 Jan. 2020, www.nytimes.com/interactive/2020/world/coronavirus-
maps.html?pageType=LegacyCollection.
4. Rabin, Roni Caryn. “Will Protests Set Off a Second Viral Wave?” The New York Times, The New York Times, 31 May 2020, www.nytimes.com/2020/05/31/health/protests-coronavirus.html.
5. Sault, Samantha, Writer. “'We're Not Ready for the next Epidemic,' Bill Gates Warned – Five Years Ago.” World Economic Forum, www.weforum.org/agenda/2020/03/bill-gates-epidemic-
pandemic-preparedness-ebola-covid-19/.
6. Stewart, Matthew, PhD Researcher. “Simple Introduction to Convolutional Neural Networks.” Medium, Towards Data Science, 29 July 2020, towardsdatascience.com/simple-introduction-
to-convolutional-neural-networks-cdf8d3077bac.
7. Mueller, John Paul. “Distinguishing Classification Tasks with Convolutional Neural Networks.” Dummies, 2 January 2019, www.dummies.com/programming/big-data/data-
science/distinguishing-classification-tasks-with-convolutional-neural-networks/.
8. American Enterprise Institute. “COVID-19: Documenting Changes in Public Opinion.” American Enterprise Institute, 2020, www.jstor.org/stable/resrep25364. Accessed 18 Oct. 2020.
9. Padan, Carmit. “Coronavirus: On Crisis, Emergency, and the Power of Words." Institute for National Security Studies, 2020, www.jstor.org/stable/resrep25527. Accessed 18 Oct. 2020.
10. Dennis, Brady, et al. “With Coronavirus Cases Spiking Nationwide, All Signs Point to a Harrowing Autumn.” The Washington Post, WP Company, 11 Nov. 2020,
www.washingtonpost.com/health/with-coronavirus-cases-spiking-nationwide-all-signs-point-to-a-harrowing-autumn/2020/11/10/d61fa050-238b-11eb-a688-5298ad5d580a_story.html.
RESOURCES

More Related Content

Similar to Are You Social Distancing? A Computational Approach

Skin cure an innovative smart phone based application to assist in melanoma e...
Skin cure an innovative smart phone based application to assist in melanoma e...Skin cure an innovative smart phone based application to assist in melanoma e...
Skin cure an innovative smart phone based application to assist in melanoma e...
sipij
 
Skin Cancer Detection and Classification
Skin Cancer Detection and ClassificationSkin Cancer Detection and Classification
Skin Cancer Detection and Classification
Dr. Amarjeet Singh
 

Similar to Are You Social Distancing? A Computational Approach (20)

Skin cure an innovative smart phone based application to assist in melanoma e...
Skin cure an innovative smart phone based application to assist in melanoma e...Skin cure an innovative smart phone based application to assist in melanoma e...
Skin cure an innovative smart phone based application to assist in melanoma e...
 
mnask detetor ppt using deep learning.pptx
mnask detetor ppt using  deep learning.pptxmnask detetor ppt using  deep learning.pptx
mnask detetor ppt using deep learning.pptx
 
IRJET - Classification of Cancer Images using Deep Learning
IRJET -  	  Classification of Cancer Images using Deep LearningIRJET -  	  Classification of Cancer Images using Deep Learning
IRJET - Classification of Cancer Images using Deep Learning
 
Face Mask Detection And Attendance System
Face Mask Detection And Attendance SystemFace Mask Detection And Attendance System
Face Mask Detection And Attendance System
 
Skin Cancer Detection and Classification
Skin Cancer Detection and ClassificationSkin Cancer Detection and Classification
Skin Cancer Detection and Classification
 
Facemask_project (1).pptx
Facemask_project (1).pptxFacemask_project (1).pptx
Facemask_project (1).pptx
 
Face Mask Detection group 14.pptx
Face Mask Detection group 14.pptxFace Mask Detection group 14.pptx
Face Mask Detection group 14.pptx
 
Automated Screening System for Acute Skin Cancer Detection Using Neural Netwo...
Automated Screening System for Acute Skin Cancer Detection Using Neural Netwo...Automated Screening System for Acute Skin Cancer Detection Using Neural Netwo...
Automated Screening System for Acute Skin Cancer Detection Using Neural Netwo...
 
Detecting anomalies in security cameras with 3D-convolutional neural network ...
Detecting anomalies in security cameras with 3D-convolutional neural network ...Detecting anomalies in security cameras with 3D-convolutional neural network ...
Detecting anomalies in security cameras with 3D-convolutional neural network ...
 
FACEMASK AND PHYSICAL DISTANCING DETECTION USING TRANSFER LEARNING TECHNIQUE
FACEMASK AND PHYSICAL DISTANCING DETECTION USING TRANSFER LEARNING TECHNIQUEFACEMASK AND PHYSICAL DISTANCING DETECTION USING TRANSFER LEARNING TECHNIQUE
FACEMASK AND PHYSICAL DISTANCING DETECTION USING TRANSFER LEARNING TECHNIQUE
 
Age and Gender Classification using Convolutional Neural Network
Age and Gender Classification using Convolutional Neural NetworkAge and Gender Classification using Convolutional Neural Network
Age and Gender Classification using Convolutional Neural Network
 
FACE_MASK.pptx
FACE_MASK.pptxFACE_MASK.pptx
FACE_MASK.pptx
 
Gesture detection in real time that serve as feedback to improve the user exp...
Gesture detection in real time that serve as feedback to improve the user exp...Gesture detection in real time that serve as feedback to improve the user exp...
Gesture detection in real time that serve as feedback to improve the user exp...
 
A State-of-the-art Review on Dielectric fluid in Electric Discharge Machining...
A State-of-the-art Review on Dielectric fluid in Electric Discharge Machining...A State-of-the-art Review on Dielectric fluid in Electric Discharge Machining...
A State-of-the-art Review on Dielectric fluid in Electric Discharge Machining...
 
Non-Invasive ABCD Monitoring of Malignant Melanoma Using Image Processing in ...
Non-Invasive ABCD Monitoring of Malignant Melanoma Using Image Processing in ...Non-Invasive ABCD Monitoring of Malignant Melanoma Using Image Processing in ...
Non-Invasive ABCD Monitoring of Malignant Melanoma Using Image Processing in ...
 
Research Paper.docx
Research Paper.docxResearch Paper.docx
Research Paper.docx
 
Monitoring Pandemic Precautionary Protocols using Real-time Surveillance and ...
Monitoring Pandemic Precautionary Protocols using Real-time Surveillance and ...Monitoring Pandemic Precautionary Protocols using Real-time Surveillance and ...
Monitoring Pandemic Precautionary Protocols using Real-time Surveillance and ...
 
FACE MASK DETECTION MODEL USING CONVOLUTIONAL NEURAL NETWORK
FACE MASK DETECTION MODEL USING CONVOLUTIONAL NEURAL NETWORKFACE MASK DETECTION MODEL USING CONVOLUTIONAL NEURAL NETWORK
FACE MASK DETECTION MODEL USING CONVOLUTIONAL NEURAL NETWORK
 
Melanoma Skin Cancer Detection using Deep Learning
Melanoma Skin Cancer Detection using Deep LearningMelanoma Skin Cancer Detection using Deep Learning
Melanoma Skin Cancer Detection using Deep Learning
 
IRJET - Body Posture Guiding System
IRJET - Body Posture Guiding SystemIRJET - Body Posture Guiding System
IRJET - Body Posture Guiding System
 

Recently uploaded

Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Safe Software
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire business
panagenda
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Safe Software
 
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Victor Rentea
 

Recently uploaded (20)

Ransomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdfRansomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdf
 
Spring Boot vs Quarkus the ultimate battle - DevoxxUK
Spring Boot vs Quarkus the ultimate battle - DevoxxUKSpring Boot vs Quarkus the ultimate battle - DevoxxUK
Spring Boot vs Quarkus the ultimate battle - DevoxxUK
 
Manulife - Insurer Transformation Award 2024
Manulife - Insurer Transformation Award 2024Manulife - Insurer Transformation Award 2024
Manulife - Insurer Transformation Award 2024
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challenges
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
 
DBX First Quarter 2024 Investor Presentation
DBX First Quarter 2024 Investor PresentationDBX First Quarter 2024 Investor Presentation
DBX First Quarter 2024 Investor Presentation
 
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire business
 
Navigating the Deluge_ Dubai Floods and the Resilience of Dubai International...
Navigating the Deluge_ Dubai Floods and the Resilience of Dubai International...Navigating the Deluge_ Dubai Floods and the Resilience of Dubai International...
Navigating the Deluge_ Dubai Floods and the Resilience of Dubai International...
 
AWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of Terraform
 
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers:  A Deep Dive into Serverless Spatial Data and FMECloud Frontiers:  A Deep Dive into Serverless Spatial Data and FME
Cloud Frontiers: A Deep Dive into Serverless Spatial Data and FME
 
MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
Artificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : UncertaintyArtificial Intelligence Chap.5 : Uncertainty
Artificial Intelligence Chap.5 : Uncertainty
 
MS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectorsMS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectors
 
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
Modular Monolith - a Practical Alternative to Microservices @ Devoxx UK 2024
 
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemkeProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
 
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
 

Are You Social Distancing? A Computational Approach

  • 1. By Aparnaa Senthilnathan, High School Senior, Bronx Science AreYouSocial Distancing?A Computational Approach
  • 2. “We’ve actually invested very little in a system to stop an epidemic...The problem wasn’t that there was a system that didn’t work well enough. The problem was that we didn’t have a system at all” —BillGates ”
  • 4. Research Goals Successfully apply mask recognition to videos Use ML algorithms to influence effective measures to stopping the pandemic Use as an analytical tool for calculating the percentage of masks Implement the program as a security measure in buildings
  • 5. If CNN models are trained on a few mask and no mask datasets and then tested on separate images and videos… then: • It will correctly identify masks and an alarm will sound that will remind non-mask wearers. • This program will correctly predict whether people are six feet or more apart. • This method will be more effective in raising awareness towards pandemics and will be ready for scientists and researchers to use the program as both an analytical tool and a security measure in buildings. Hypothesis
  • 6. Methods 02. CNN Models, First Times?, Equations, Main Menu Module, Details
  • 7. ● Neural Networks ● Used to train the CNN model to identify items in images ● Consists of four layers: convolutional, pooling, fully connected and softmax. ● These layers are ordered by how they learn filters of increasing complexity. CNNModels
  • 8. • Use machine learning concepts to create a program to help encourage more people to follow the guidelines necessary for Coronavirus prevention • Attempt to implement mask recognition algorithms into video recordings. Firsttimes?
  • 9. Materials The program was coded using Python and its libraries. A few mask and no mask datasets from Kaggle and Github were download.
  • 10. TSC is the number of test samples the model was able to correctly identify and the TS is the total number of test samples received by the model. The loss is just calculated by subtracting the accuracy from 1, where 1 is the maximum percentage of accuracy Accuracy Loss Equations A(TSC,TS) = TSC/TS L(A) = 1 - A
  • 11. P is the percentage of people wearing masks, M is the number of masks identified, and NP is the number of people in the frame. Percentageofmasks Equations(cont.) P = M/NP
  • 12. TheProgram’s mainMenuModule Take a Picture using the Computer Open a Webcam Video on your Computer Upload an Image Upload a Video 01 02 03 04 Exit 05
  • 13. Demo The photo sent by the user Percentage of mask- wearers over a total number of people A green border indicates the person is wearing a mask If the person isn’t wearing the mask, then the border should be red.
  • 15. Table1:TakeaPhoto Trials(MaskPercentage/#ofPeoplerecorded/TotalFacesshown) Photo Type Trial 1 Trial 2 Trial 3 Single Person (Mask) 0%/1(plastic mask) 100%/1 (pink mask) 100%/1 (black mask) Single Person (No-Mask) 0%/1 0%/1 0%/1 Two People (Both Masks) 50%/2 100%/2 50%/2 (-wrong position) Two People (One Mask) 50%/2 (wrong face) 100%/2 (person without a mask counted) 50%/2 Two People (No Masks) 0%/2/2 0%/2/2 0%/2/2 Group of People (All Masks) 66.7%/2/3 100%/3/3 33.3%/1/3 Group of People (Some Masks) 66.7%/3/3 66.7%/3/3 33.3%/1/3 Group of People (No Masks) 33.3%/3/3 (one non-mask wearer counted) 33.3%/3/3 (one non-mask wearer counted) 33.3%/3/3 (one non-mask wearer counted)
  • 16. Table2:UploadaPhoto Trials(MaskPercentage/DistanceAccuracy/#ofPeoplerecorded/Total Facesshown) Photo Type Trial 1 Trial 2 Trial 3 Single Person (Mask) 100%/100%/2 (pillow counted/1) 100%/100%/1/1 100%/100%/1/1 Single Person (No-Mask) 0%/100%/1/2 0%/100%/1/2 0%/100%/1/2 Two People (Both Masks) 100%/100%/2/2 100%/100%/2/2 100%/100%/2/2 Two People (One Mask) 100%/100%/1/2 0%/0%/0/2 50%/100%/2/2 Two People (No Masks) 100%/100%/2/2 50%/100%/2/2 50%/100%/2/2 Group of People (All Masks) 0%/100%/2/4 75%/50%/8/13 (girl’s shoulder counted) 86.66%/100%/15 /50 Group of People (Some Masks) 100%/0%/5/20 (a non-mask wearer counted) 53.85%/100%/13 /13 (phone counted) 70%/100%/28/30 (few non-mask wearers counted) Group of People (No Masks) 0%/0%/10/10 0%/0%/9/9 0%/0%/6/7
  • 17. Table3:Webcam Trials(MaskPercentage/PeopleIdentified/Numberof People) Photo Type Trial 1 Single Person (Mask) 0%/1/1 Single Person (No-Mask) 100%/1/1 Two People (Both Masks) 50%/2/2 Two People (One Mask) 50%/2/2 Two People (No Masks) 0%/2/2 Group of People (All Masks) 33.3%/1/3 Group of People (Some Masks) 66.7%/2/3 Group of People (No Masks) 0%/3/3
  • 18. Table4:Upload aVideo VideoName Observations How to wear your single use face mask? (short version) Before the girl puts on her mask, the program identifies that her face has a mask on. As the girl was putting on her mask, the camera turned 360 degrees in slow motion, which the program wasn’t able to identify her until her face was shown in front view. The program was able to identify her with the mask on correctly, but without the mask, the program still identifies her as her wearing the mask. Group of People Putting on Masks Three people are recorded in front view by the camera, and the program was able to classify which of the three people are wearing masks as well as the distance between them for the most part.
  • 19. dISCUSSION What can we infer?/Implications 04.
  • 20. Observations Photos ● There were mixed results ● The program was able to identify and classify masks correctly for the most part. ● The program also calculated the percentage of masks within the image correctly. ● In some cases, the program can’t identify other faces due to lighting conditions or the person is wearing the mask Videos ● There also were mixed results ● The program can identify and classify the face properly when the subject is facing the camera. Otherwise, it is hard for the program to identify the face. ● The program also calculated the percentage of masks within the video correctly. ● The alarm is correctly initiated when the program identified a non-mask wearer in the webcam video.
  • 21. Mask recognition can be used successfully in videos to an extent For the most part, the program can serve as an analytical tool because it calculates the percentage of masks correctly. VideoRecognition AnalyticalTool Discoveries
  • 22. The program allowed the pre- trained model to identify an object as a face, despite a low confidence score. If the video size is large, the video playback would slow down a lot due to the excessively large amount of data the model has to read. Wrongface Videoprocessing Limitations No mask
  • 23. • Find ways to improve the machine learning algorithms such that the program can process videos in original sizes and allow the program to identify the right faces in the same video. • Can be expanded further by training the model to identify whether or not the person is wearing the mask correctly and is it the right type of mask. • Introduce future methodology that allow scientists to jump further into raising awareness towards the coronavirus pandemic. Future RESEarch
  • 24. • All icons and images are created by Flaticon and Freepick. The screenshots of code/demo are from my laptop. The presentation template was created by Slidesgo. • Special thanks to Dr. Sarah Ostadabbas of Northeastern University and members from the MIT Beaver Works Summer Institute for mentoring and helping me with CNN Networks and resources necessary for the project. • Also, a big thanks to Dr. Vladimir Shapovalov for guiding me through the research process Acknowledgements
  • 25. 1. The University of Texas at Austin. "Coronavirus spreads quickly and sometimes before people have symptoms, study finds." ScienceDaily. ScienceDaily, 16 March 2020. <www.sciencedaily.com/releases/2020/03/200316143313.htm>. 2. Bai, Nina. “How the New Coronavirus Spreads and Progresses – And Why One Test May Not Be Enough.” How the New Coronavirus Spreads and Progresses – And Why One Test May Not Be Enough | UC San Francisco, 10 Nov. 2020, www.ucsf.edu/news/2020/02/416671/how-new-coronavirus-spreads-and-progresses-and-why-one-test-may-not-be-enough. 3. The New York Times. “Covid World Map: Tracking the Global Outbreak.” The New York Times, The New York Times, 28 Jan. 2020, www.nytimes.com/interactive/2020/world/coronavirus- maps.html?pageType=LegacyCollection. 4. Rabin, Roni Caryn. “Will Protests Set Off a Second Viral Wave?” The New York Times, The New York Times, 31 May 2020, www.nytimes.com/2020/05/31/health/protests-coronavirus.html. 5. Sault, Samantha, Writer. “'We're Not Ready for the next Epidemic,' Bill Gates Warned – Five Years Ago.” World Economic Forum, www.weforum.org/agenda/2020/03/bill-gates-epidemic- pandemic-preparedness-ebola-covid-19/. 6. Stewart, Matthew, PhD Researcher. “Simple Introduction to Convolutional Neural Networks.” Medium, Towards Data Science, 29 July 2020, towardsdatascience.com/simple-introduction- to-convolutional-neural-networks-cdf8d3077bac. 7. Mueller, John Paul. “Distinguishing Classification Tasks with Convolutional Neural Networks.” Dummies, 2 January 2019, www.dummies.com/programming/big-data/data- science/distinguishing-classification-tasks-with-convolutional-neural-networks/. 8. American Enterprise Institute. “COVID-19: Documenting Changes in Public Opinion.” American Enterprise Institute, 2020, www.jstor.org/stable/resrep25364. Accessed 18 Oct. 2020. 9. Padan, Carmit. “Coronavirus: On Crisis, Emergency, and the Power of Words." Institute for National Security Studies, 2020, www.jstor.org/stable/resrep25527. Accessed 18 Oct. 2020. 10. Dennis, Brady, et al. “With Coronavirus Cases Spiking Nationwide, All Signs Point to a Harrowing Autumn.” The Washington Post, WP Company, 11 Nov. 2020, www.washingtonpost.com/health/with-coronavirus-cases-spiking-nationwide-all-signs-point-to-a-harrowing-autumn/2020/11/10/d61fa050-238b-11eb-a688-5298ad5d580a_story.html. RESOURCES