Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Mask Wearing Detection Using OpenCV Training Data

Mask Wearing Detection Using OpenCV Training Data

https://aaron.kr/content/portfolio/can-we-detect-mask-wearing-with-a-pc-webcam/

Poster presentation at The 50th Fall Comprehensive Conference of the Korea Information and Communication Society (KIICE) (제50회 한국정보통신학회 추계종합학술대회).

Using haarcascade XML files that can automatically detect faces, eyes, mouths, and noses, we perform a simple mask wearing detection using a common PC webcam.

Aaron Snowberger

October 28, 2021
Tweet

More Decks by Aaron Snowberger

Other Decks in Technology

Transcript

  1. OpenCV 학습 데이터를 이용한 마스크 착용 감지
    Mask Wearing Detection
    Using OpenCV Training Data
    Aaron Daniel Snowberger,
    Choong Ho Lee
    Mask
    No Mask

    View full-size slide

  2. INTRODUCTION
    It is an important issue to detect automatically
    whether a mask is worn or not for corona
    prevention. It is known that mask wearing
    detection can be solved by learning the face
    data set. However, the search for whether a
    person is wearing a mask can be detected in a
    simpler way using OpenCV. In this paper, we
    describe that it is possible to easily detect
    whether a single person is wearing a mask or
    not with a general PC camera using OpenCV
    learning data results and simple OpenCV
    functions. Through experiments, the proposed
    method was shown to be effective.

    View full-size slide

  3. BUILT WITH
    PYTHON
    3.9.7
    OPENCV
    4.5.1
    TENSORFLOW
    2.5.1
    haarcascade_frontalface_default.xml
    haarcascade_eye.xml
    haarcascade_mcs_mouth.xml
    haarcascade_mcs_nose.xml

    View full-size slide

  4. FIRST, FIND THE FACE AREA
    In the upper half
    of the face area
    EYES
    In the lower half
    of the face area
    MOUTH
    In the vertical ⅓ ~ ⅔
    of the face area
    NOSE

    View full-size slide

  5. IMPLEMENTATION & RESULTS
    WITH MASK
    WITHOUT MASK
    If eyes, nose, and mouth
    are found, the mask is not worn
    If nose and mouth are not found,
    it is judged that the mask is worn
    No Mask Mask

    View full-size slide

  6. EXTENDED EXPERIMENTATION
    Camera angle & lighting
    matters. Some mistakes in
    suboptimal conditions.
    01Angle / light
    02
    It’s possible to detect
    partially worn masks if
    nose or mouth is detected.
    03
    It’s possible to trick the
    camera by obscuring the
    nose & mouth.
    COVER it up
    Chin diaper

    View full-size slide

  7. CONCLUSION
    Instead of re-learning with a
    mask-wearing face, we implemented a
    method of retrieving a mask when the
    nose and mouth are detected in the face
    area with the previously learned results,
    and that the nose and mouth are not
    searched at the same time. There
    remains a need to conduct research on
    whether similar results can be obtained
    by running this program on Raspberry Pi
    to implement a device for checking
    standing masks, or by using CCTV
    instead of a PC camera.

    View full-size slide

  8. REFERENCES
    [1] Dong-Guen Kim, OpenCV Programming
    with Python, Kame Press, 2018.
    [2] Youtube. Artificial intelligence that
    detects whether a mask is on or not
    [Internet] Available:
    https://www.youtube.com/watch?v=ncIyy1
    doSJ8&t=66s
    [3] Github. [Internet] Available:
    https://github.com/peterbraden/node-open
    cv/tree/master/data
    Mask
    CREDITS: This presentation template was created by Slidesgo,
    including icons by Flaticon, infographics & images by Freepik

    View full-size slide