The pandemic has made face masks essential in our each day lives. As increasingly countries mandate using face masks to regulate the spread of the virus, it has turn out to be increasingly necessary to make sure that people comply with these regulations. One method to do that is through the use of face mask detection technology.

On this blog, we’ll explore detect whether an individual is wearing a face mask using and library.

To start with, we want to know what’s Algorithm. It’s a light-weight yet powerful machine-learning algorithm used for solving each classification-type and regression-type problems.

The algorithm finds the very best decision boundary that separates the information points into different classes.

1. Data points are plotted on a graph, where each point represents an remark with several features.
2. The algorithm tries to search out a line (or a hyperplane in higher dimensions) that separates the 2 classes with the biggest possible margin.
3. The margin is the gap between the choice boundary and the closest data points from each classes.
4. The info points closest to the choice boundary are called support vectors, they usually determine the position and orientation of the boundary.
5. Once the boundary is found, the algorithm can classify latest data points by determining on which side of the boundary they fall.

SVMs may also be used for non-linear classification tasks through the use of kernel functions to remodel the information into the next dimensional space where a linear boundary could be found.

Now we will probably be implementing this into our project. We will simply define this as a where we are able to find if an individual is wearing a mask or not.

The algorithm is defined within the Spackage. is a module for machine learning. seamlessly quickens your scikit-learn applications for and across single- and multi-node configurations. Acceleration is achieved through using the . Intel(R) Extension for Scikit-learn incorporates scikit-learn patching functionality that was originally available within the package. This extension package dynamically patches estimators while improving the performance of your machine-learning algorithms.

The extension is an element of the Intel® AI Analytics Toolkit (AI Kit) that gives flexibility to make use of machine learning tools along with your existing AI packages. Using with this extension, we are able to speed up training and inference by as much as with the equivalent mathematical accuracy.

• Python 3.7 or above.
• Packages Required : Numpy, Pandas, scikit-learn-intelex, scikit-learn, PIL, matplotlib

You’ll be able to create a virtual environment to put in these packages. Write down the next code on a cell of the Jupyter Notebook:

!pip install scikit-learn-intelex

Then run:

from sklearnex import patch_sklearn
patch_sklearn()

The dataset has been taken from . It incorporates photos with two different labels, one is and the opposite one is With_mask dataset incorporates 3725 images and the Without_mask dataset incorporates 3828 images.

Dataset Link: https://www.kaggle.com/datasets/omkargurav/face-mask-dataset

Used PIL and Numpy libraries to import the dataset after which convert it into 3 channel numpy array.

with_mask_files = os.listdir(‘data/with_mask’)
print(with_mask_files[0:5])
print(with_mask_files[-5:])

without_mask_files = os.listdir(‘data/without_mask’)
print(without_mask_files[0:5])
print(without_mask_files[-5:])

Now we’ll store the Convert it into Numpy files

The pictures at the moment are converted into 3 channel arrays, each element of the array incorporates 3 parameters of the image().

Now before further moving on to model training, we want to alter the scale of the array from 3 channels to a single channel.

Now we will probably be appending all of the values right into a single list, creating one other list of labels containing 0s, 1s and will probably be using train_test_split. to separate the dataset into training and testing.

We will probably be using the support vector classifier of SVM to suit our model after which make predictions.

Now we are able to make predictions and check the accuracy of our model.

Checking the :

Now we are able to test our model with some images if our model is working properly or not.

Let’s check one other image of an individual with out a mask.

This model may also be implemented right into a computer vision model which may detect masks on a face in real-time.

This project was showcased by (Myself) at organized by in partnership with , , and .

Thanks on your time.