Face Mask Detection in Images Through Deep Learning
Disciplines
Artificial Intelligence and Robotics | Programming Languages and Compilers
Abstract (300 words maximum)
Since the beginning of the Covid-19 pandemic, a highly infectious respiratory illness, across the world face masks have risen in popularity as an effective method to mitigate the transmission of the diseases. The heightened awareness of face masks has brought awareness to the effectiveness of enforcing mask policies past the mandatory face covering rules in public places of high transmission, such as in hospitals and assisted living facilities [1] due to high crowd densities or higher than average infected individuals. The increased number of individuals continuing to wear masks, as well as the stricter enforcement of mask polices in areas of high transmission have necessitated the further investigation and evaluation of mask detection systems to further curb the transmission of Covid-19 and any future pandemic. Faster and more efficient means of classifying faces wearing masks holds significant value in edge devices and isolated security systems. In this paper we conduct a review on several of the most popular image classification methods being used today and run a test to see how well they can classify human faces as “mask is present” or “mask is not present” in comparison to a new model designed by the authors of this paper. Additionally, based on the models’ accuracy and efficiency, we propose which technique is best and should be used for those who seeking to implement an accurate and fast mask-detection model for real world settings.
Academic department under which the project should be listed
CCSE - Computer Science
Primary Investigator (PI) Name
Md Abdullah Al Hafiz Khan
Face Mask Detection in Images Through Deep Learning
Since the beginning of the Covid-19 pandemic, a highly infectious respiratory illness, across the world face masks have risen in popularity as an effective method to mitigate the transmission of the diseases. The heightened awareness of face masks has brought awareness to the effectiveness of enforcing mask policies past the mandatory face covering rules in public places of high transmission, such as in hospitals and assisted living facilities [1] due to high crowd densities or higher than average infected individuals. The increased number of individuals continuing to wear masks, as well as the stricter enforcement of mask polices in areas of high transmission have necessitated the further investigation and evaluation of mask detection systems to further curb the transmission of Covid-19 and any future pandemic. Faster and more efficient means of classifying faces wearing masks holds significant value in edge devices and isolated security systems. In this paper we conduct a review on several of the most popular image classification methods being used today and run a test to see how well they can classify human faces as “mask is present” or “mask is not present” in comparison to a new model designed by the authors of this paper. Additionally, based on the models’ accuracy and efficiency, we propose which technique is best and should be used for those who seeking to implement an accurate and fast mask-detection model for real world settings.