Thesis
Enhancing face mask detection using deep neural networks and transfer learning for COVID-19 transmission prevention
University of Iowa
Master of Science (MS), University of Iowa
Spring 2021
DOI: 10.17077/etd.005925
Abstract
Artificial Intelligence and Deep Learning have proven to be very effective in helping to learn about and prevent the spread of COVID-19 since the start of the global pandemic. One area that they have been prominent alongside Computer Vision is in the field of mask detection and classification. COVID-19 has been determined to be a virus that is transmitted primarily through particles that leave the nose and mouth. The CDC has suggested that masks and other face coverings can help to stop the transmission of the virus. The CDC, as well as several other sources, note that for masks to be effective, they must be worn securely over both the nose and mouth. Using different surveillance techniques to detect and classify masks allows for the enforcement of mask-wearing as well as the ability to understand mask-wearing trends. Previous works have focused only on whether or not a person is wearing a mask and some attempt to classify if a person is wearing the mask correctly or not. This thesis explores an approach to detection using Multi-Output Deep Transfer Learning models to enhance current detection and classification techniques. The added outputs of this model include detecting whether a person is wearing the mask over the chin or with the nose exposed in the case of incorrect wearing. The Multi-Output model achieves an average accuracy of over 97% on a combination of real and generated data and proves to be an enhancement to the previous works. This model will serve as a crucial building block for a larger mask detection system in efforts to slow the spread of the virus.
Details
- Title: Subtitle
- Enhancing face mask detection using deep neural networks and transfer learning for COVID-19 transmission prevention
- Creators
- Austin Strom
- Contributors
- Guadalupe Canahuate (Advisor)Alberto Segre (Committee Member)Ted Herman (Committee Member)
- Resource Type
- Thesis
- Degree Awarded
- Master of Science (MS), University of Iowa
- Degree in
- Electrical and Computer Engineering
- Date degree season
- Spring 2021
- DOI
- 10.17077/etd.005925
- Publisher
- University of Iowa
- Number of pages
- ix, 34 pages
- Copyright
- Copyright 2021 Austin Strom
- Language
- English
- Description illustrations
- color illustrations
- Description bibliographic
- Includes bibliographical references (pages 31-34)
- Public Abstract (ETD)
Artificial Intelligence and Deep Learning have proven to be very effective in helping to learn about and prevent the spread of COVID-19 since the start of the global pandemic. One area that they have been prominent alongside Computer Vision is in the field of mask detection and classification. Using different surveillance techniques to detect and classify masks allows for the enforcement of mask-wearing as well as the ability to understand mask wearing trends. This thesis explores an approach to detection using Multi-Output Deep Transfer Learning models to enhance current detection and classification techniques. The Multi-Output model achieves an average accuracy of over 97% on a combination of real and generated data and proves to be an enhancement to the previous works.
- Academic Unit
- Electrical and Computer Engineering
- Record Identifier
- 9984097073902771
Metrics
10 File views/ downloads
351 Record Views