IRJMETS Sri Ram.j

e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science

Volume:03/Issue:03/March-2021 Impact Factor- 5.354 www.irjmets.com
DETECTING FACEMASK USING DEEP LEARNING
Jonnalagadda Sri Ram*1, DR. Vivek Duraiswamy Maikandavel*2
*1
, ECE, SRM IST, Chennai, Tamil Nadu, India.
(*2Research Assistant Professor, ECE, SRM IST, Chennai, Tamil Nadu, India)
ABSTRACT
Even while vaccines are being rolled out for novel coronavirus, it doesn't mean to stop using facemasks. As per the
latest reports from WHO (World Health Organization), new mutation variants of coronavirus are spreading even
faster. At present many governments are re-imposing lockdowns even with economic backlash situations. Using
sanitizers, facemasks, and maintaining social distancing is very important in the present scenario. With a risk-free
and cost-effective approach by using AI, we can create a secure environment. The research uses deep learning
strategies to differentiate facial recognition and to recognize whether or not a person is wearing a face mask. We're
going to use OpenCV to render real-time face recognition from a live stream. This research is useful in the fight
against the transmission of the virus and in preventing contact with the virus.
Keywords: OpenCV, Machine learning, Deep learning, Novel Corona Virus Disease – 2019,
I. INTRODUCTION
According to recent reports, COVID-19 has affected more than 11.4 crore people worldwide, resulting in more than
25.3 lakh deaths. This coronavirus outbreak has given rise to an unprecedented degree of scientific collaboration
throughout the world. Artificial Intelligence (AI) will help combat Covid-19 in several respects. Machine learning (ML)
helps researchers and practitioners to examine large volumes of data to simulate the spread of COVID-19, to act as
an advance warning system for future pandemics, and to identify susceptible populations. In present situations, the
legislation obliges citizens to wear facial masks in public in several nations. These policies and regulations have been
developed as an intervention for rapid increases of cases and deaths in many places. Even so, the task of tracking
large numbers of individuals is becoming more challenging. The monitoring process includes the identification of
someone who does not wear a face mask. In this analysis, deep learning methods are used to create a classifier that
extracts photographs of a human wearing a mask and not from a dataset and distinguishes between these classes of
face masking and not face masking. The artificial neural network (ANN) has been shown to be a robust procedure for
the extraction of features from unprocessed results. The OpenCV, TensorFlow, and Python programming languages
were used to build this project.
II. LITERATURE REVIEW
[1] In this work, a transfer learning-based method for classifying people who aren't wearing face masks is proposed,
which makes use of the InceptionV3 pre-trained model. The very last layer of the InceptionV3 is withdrawn and the
network is finely tuned by the addition of 5 more layers. [2] Here, To boost model efficiency without overfitting the
training phase, the last layer in ResNet-50 was removed and replaced with three conventional machine learning
classifiers (Support vector machine, ensemble, and decision tree). [3] a model has been created to classify facial
images using CNN architecture and used deep learning methods (Artificial neural networks) for identifying whether
or not facemask is used by a person, Finally deployed in raspberry pi. [4] FACEMASKNET was created using a
convolutional neural network framework that was specifically used for learning and interpreting images. To perfect
the learning result, the teaching pedagogy used mathematical algorithms (MATLAB software) and several
comparative relearning steps (layers).
III.METHODOLOGY
The developed model focuses on how to identify a person wearing a face mask on a picture/video transfer using
deep learning calculations using OpenCV and Tensorflow. The dataset has been augmented into two parts as “mask”
and “no mask” and trained.
Data pre-processing
1.The input frame is adjusted (100 x 100).
2. Color filtering is applied to the streams.
3. Using the regular mean of PyTorch's built-in weights, frames are scaled and normalized.
4. Cropping the frame in the middle with a pixel value of 200x200x3.
5. Similar to NumPy array they are transformed to tensors.
www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1]
e-ISSN: 2582-5208
Figure 1: Model flow chart.

When a frame is processed and if it is needed evaluation of mask or without mask, a new thread will start as if we
used multi-threading.
IV.RESULTS AND DISCUSSION

By training the model 99% accuracy has been achieved. The below screenshots in fig 2,3 are the highest recorded
outputs yet after several attempts conducted for face mask detection. Whereas fig 2 is output for the photos
uploaded while fig 3 is the output from the video stream. In fig 2 the left image is the result from the image uploaded
with facemask and the accuracy rate is about 99.7% where the right image is the output without facemask with an
accuracy of 97.2%. While in fig 3 the left images are the output from the video stream with face mask with an
accuracy rate of 98.5% and the right image is the output for without face mask with an accuracy rate of 98.7%.
Figure 2: Uploded images output.
Figure 3: video output.

[2]
e-ISSN: 2582-5208
V.CONCLUSION
Many previous studies have looked into the method for face detection in the wild, and the related face detectors
have been checked on datasets of normal faces. Some face detectors have obtained exceptionally superior efficiency
on these datasets, and it appears to be challenging to enhance them further. The real-world situations, on the other
hand, are much more difficult than predicted in terms of containing faces captured at unusual resolutions, lighting,
and compression. The trained model is able to complete its task using the model architecture, achieving an output
accuracy of 99 percent.
This very same study can be enhanced further by using vast quantities of data, and it can also be applied to identify
the form of a mask and incorporate a facial recognition device, which can be implemented at different workplaces to
aid in individual identification when wearing the mask.
VI.REFERENCES
[1] Jignesh Chowdary G., Punn N.S., Sonbhadra S.K., Agarwal S. (2020) Face Mask Detection Using Transfer Learning
of InceptionV3. In: Bellatreche L., Goyal V., Fujita H., Mondal A., Reddy P.K. (eds) Big Data Analytics. BDA 2020.
Lecture Notes in Computer Science, vol 12581. Springer, Cham. https://doi.org/10.1007/978-3-030-66665-1_6
[2] Mohamed Loey, Gunasekaran Manogaran, Mohamed Hamed N. Taha, Nour Eldeen M. Khalifa, A hybrid deep
transfer learning model with machine learning methods for face mask detection in the era of the COVID-19
pandemic, Measurement, Volume 167,2021,108288, ISSN 0263-2241.
[3] S. V. Militante and N. V. Dionisio, "Real-Time Facemask Recognition with Alarm System using Deep Learning,"
2020 11th IEEE Control and System Graduate Research Colloquium (ICSGRC), Shah Alam, Malaysia, 2020, pp.
106-110, doi:10.1109/ICSGRC49013.2020.9232610.
[4] Inamdar, Madhura and Mehendale, Ninad, Real-Time Face Mask Identification Using Facemasknet Deep
Learning Network (July 29, 2020). Available at SSRN: https://ssrn.com/abstract=3663305
[5] DS, Karthik Srivathsa, A. Rengarajan, and Nikhil Kumar HS. DETECTING OF FACE MASK.
Volume:02/Issue:12/December -2020.
[6] Vinitha.V1, Velantina.V2. COVID-19 FACEMASK DETECTION WITH DEEP LEARNING AND COMPUTER VISION.
Volume: 07 Issue: 08, Aug 2020.
[7] S. Ge, J. Li, Q. Ye and Z. Luo, "Detecting Masked Faces in the Wild with LLE-CNNs," 2017 IEEE Conference on
Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017, pp. 426-434, doi:
10.1109/CVPR.2017.53.

[3]

IRJMETS Sri Ram.j

Uploaded by

Copyright:

Available Formats

IRJMETS Sri Ram.j

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

IRJMETS Sri Ram.j

Uploaded by

Copyright:

Available Formats

e-ISSN: 2582-5208

International Research Journal of Modernization in Engineering Technology and Science

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

Figure 1: Model flow chart.

IV.RESULTS AND DISCUSSION

Figure 2: Uploded images output.

Figure 3: video output.

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

You might also like