MaskSense

Inspiration

Only 44% of people in the US always practice the important safety measure of wearing a mask during the COVID-19 pandemic. This statistic is extremely low and is detrimental towards the elimination of COVID-19 in the US.

What it does

My application predicts if a user is currently wearing a face mask through their webcam real-time.

How I built it

I built a CNN Sequential model using tensorflow.keras, and trained it using a dataset from Kaggle. The dataset can be found here: https://www.kaggle.com/harry418/dataset-for-mask-detection?

I used the libraries numpy, keras, sklearn, tensorflow, and OpenCV.

Challenges I ran into

The primary challenge I ran into was playing around with the model to get the highest validation accuracy rate. After testing with different techniques and parameters such as pooling and learning rate, my model's accuracy increased significantly.

Accomplishments that I'm proud of

I'm proud of the accuracy that my model achieved; my final model reached an accuracy of 90%.

What I learned

I expanded my understanding of building ML models, as well as learned how to use features of the OpenCV library.

What's next for MaskSense

This application is relevant to the current pandemic, and is scalable for various use cases, such as at institutions and public spaces to enforce face masks.

Built With

• keras
• opencv
• python
• scikit-learn
• tensorflow