This document describes a computer vision system to automatically detect violations of face mask wearing during the COVID-19 pandemic. It uses techniques like Faster R-CNN, OpenCV for face detection, and MobileNet V2 in a TensorFlow/Keras model. The system was trained and tested on a dataset of 1376 images with 690 containing masks and 686 without masks. It achieved high accuracy in detecting masked and unmasked faces. Future work includes deploying this on edge devices like drones to monitor social distancing and mask compliance.
1. INTRODUCTION
THE COVID-19 PANDEMIC HAS CAUSED MANY SHUTDOWNS IN
DIFFERENT INDUSTRIES AROUND THE WORLD. THE WORLD
HEALTH ORGANIZATION RECOMMENDS WEARING A FACE
MASK AND PRACTICING COVID APPROPRIATE BEHAVIOR. THIS
PROJECT DEVELOPED A COMPUTER VISION SYSTEM TO
AUTOMATICALLY DETECT THE VIOLATION OF FACE MASK-
WEARING TO ASSURE THE SAFETY OF PEOPLE.
2. PROBLEM STATEMENT
• During the Pandemic period, face masks have proved
to be the most effective step toward preventing the
spread of COVID 19. However, it is difficult for an
organization to ensure this behavior on a collective
basis.
• People often violate this rule which the organization
has to suffer. like other sectors the construction
industry has been affected where many projects have
been suspended due to the pandemic. In the wake of
this, we have designed software that will solve this
3. LITERATURE REVIEW
Face mask detection identifies whether a person is wearing a
mask or not in a picture Jiang et al. [21] employed a one-stage
detector, called Retina FaceMask, that used a feature pyramid
network to fuse the high-level semantic information. They added
an attention layer to detect the face mask faster in an image.
Militante & Dionisio [24] used the VGG-16 CNN model and
achieved 96% of accuracy to detect people who wear a face mask
or not. Research studies have collected large scale image data
from drones and applied deep learning techniques to detect
objects
5. MAIN TECHNOLOGIES/LIBRARIES
USED
• FASTER R CNN
•IN FACE DETECTION- OPEN CV
•TENSORFLOW AND KERAS TO MSKE CUSTOM MASK DETECTOR
•MOBILE NET V2
•IMAGE DATA GENERATOR
•INUTILS, TIME, OS ETC
6. IMAGE PROCESSING- OPEN CV
(FACE DETECTION)
• HOW A COMPUTER SEE AN IMAGE
• •VIOLA JONES ALGORITHM
• •HAAR FEATURES
• •HAAR CASCADE MODEL USED FROM OPEN CV –
“haarcascade_frontalface_ak2.xml”
7. • MOBILE NET V2
THE MOBILENET V2 ARCHITECTURE IS BASED ON AN INVERTED RESIDUAL STRUCTURE
WHERE THE INPUT AND OUTPUT OF THE RESIDUAL BLOCK ARE THIN BOTTLENECK LAYERS
OPPOSITE TO TRADITIONAL RESIDUAL MODELS WHICH USE EXPANDED REPRESENTATIONS IN
THE INPUT
• IMAGE DATA GENERATOR
THE IMAGEDATAGENERATOR CLASS IN KERAS IS USED
FOR IMPLEMENTING IMAGE AUGMENTATION. THE MAJOR
ADVANTAGE OF THE KERAS IMAGEDATAGENERATOR CLASS IS ITS
ABILITY TO PRODUCE REAL-TIME IMAGE AUGMENTATION. THIS
SIMPLY MEANS IT CAN GENERATE AUGMENTED IMAGES
DYNAMICALLY DURING THE TRAINING OF THE MODEL MAKING THE
OVERALL MODE MORE ROBUST AND ACCURATE.
9. TRAINING AND TESTING DATA:
• Turns oodles of data into
predictions.
• Offer a lot of profit potential
• ML algorithms require quality
training and testing.
• A lot of it is also done for max
accuracy.
• Our dataset has 1376 images to
train and test.
• 690 images are with mask.
• 686 images are without mask.
10. CREATION OF FACE MASK DATASET
• Mechanism for detecting face.
• Image of person.
• Computation of boundary box
location of face.
• Extraction of face region of
interest(ROI).
• Apply facial landmarks to localise
eyes, mouth, etc.
• Mask applied to face.
11. USE OF VARIOUS LIBRARIES AND
FUNCTIONS:
• Use of tensorflow and keras.
• Fine tuning MobileNet V2.
• Set of tensorflow.keras imports.
• Use of SciKit- learn.
• Imutils path implementation.
• Plot training loss and accuracy.
17. DUE TO THE URGENCY OF CONTROLLING COVID-19, THE
APPLICATION VALUE AND IMPORTANCE OF REAL-TIME MASK
AND SOCIAL DISTANCING DETECTION ARE INCREASING.
FIRSTLY, THIS WORK REVIEWED MANY RESEARCH WORKS THAT
SEEK TO SURROUND THE COVID-19 OUTBREAK. THEN, IT
CLARIFIED THE BASIC CONCEPTS OF DEEP CNN MODELS.
IN FUTURE WORKS, WE WILL EXPLOIT THIS METHODOLOGY ON
SMART SENSORS OR CONNECTED RP NODES THAT WILL BE
CONSIDERED AS AN EDGE CLOUD TO COLLECT MULTIMEDIA
DATA, E.G., AN AUTONOMOUS DRONE SYSTEM, WHICH CAN
PROVIDE CAPTURE (BY THE CAMERA) OF THE DETECTED
OBJECTS FROM DIFFERENT ANGLES AND SEND THEM TO THE
EDGE CLOUD SYSTEM TO BE ANALYZED.
Future scope