AI Based Face Mask And Social Distancing Detection

Inspiration

During this pandemic time, the world was facing a lot of problems. Thus I wanted to contribute to the society and make this project which makes a difference in the real world. The government of India along with WHO was constantly advising to follow specific guidelines for the pandemic time. Therefore, I wanted to do my part by making this project which can help everyone in ensuring that everyone follows the safety norms.

What it does

There are two sub projects in this project: 1) Face Mask Detection- This program detects whether a person is wearing a face mask or not 2)Social Distancing Detection- This program checks whether social distancing is being followed or not

How I built it

For both the programs, I applied Transfer Learning !

*For the Mask Detection Program, I used Mobilenet SSD v2 model which was trained by Google and Face Detection Dataset from Kaggle If the mask was detected, the person's face will be surrounded with a green box having label MASK else the person will have a red box along the face with the label no mask

*For the Social Distancing Detection Program, I used the pre-trained model from tensorflow model zoo. First I classified the person, calculated the midpoints and finally found the distance between the midpoints using Euclidean Distance. Then I checked if the distance between the adjacent people is less than or equal to 100. If yes, then I changed the color of the rectangular box to red indicating social distancing not being followed

Challenges I ran into

I ran into many challenges such as which Models should I use, should I use Machine Learning or Just Deep Learning, should I use old classification methods such as HOG(Histogram Oriented Gradient) or should I try the new ones, etc. The biggest challenge I faced was when I was not able to run one of my program. But later I debugged it and was able to run it successfully.