This project was inspired by our love for machine learning as well as our passion for helping out society during an unprecedented time. Our own families are more susceptible to catching COVID-19 so we wanted to create a project which would prevent the spread of the virus as we start returning to in-person lives.

What it does

miramask connects to live video feed from cameras to monitor whether people at an establishment are wearing masks. Users can keep track of various cameras with the help of our application. The app will notify users when a person not wearing a mask is detected. We aimed to create this project primarily for healthcare facilities (such as hospitals, doctor's offices, and clinics) as well as brick-and-mortar businesses (including grocery stores, small businesses, and banks).

How we built it

We started by using Python (specifically, the Tensorflow library) as well as datasets from Kaggle (training, test) to create an ML model which can classify images based on whether a face is wearing a mask or not. Then, we connected our training model to OpenCV (Intel's open source computer vision library) and our computer webcams so we could detect faces from the live video feed and determine whether they wore masks. Finally, we created the iPhone user interface using Moqups. Users can view the live video feed from their cameras 24/7 and receive notifications when someone without a mask is detected.

Challenges we ran into

We had trouble finding suitable datasets (some were structured in ways that were not compatible with the architecture of our model; others did not have enough images). Keeping our pitch succinct yet informative was another challenge.

Accomplishments that we're proud of

Our ML model reached an accuracy of 99% and a val_accuracy of 91%! Being able to see our mask detection program working live on our computers was very rewarding and satisfying. We are really happy with how our UI looks. We're also proud of figuring out how to upload our video to Youtube :)

What we learned

We learned how to use OpenCV and Moqups! We also furthered our knowledge of Python and Tensorflow. Most importantly, we learned how to concisely and effectively present our project.

What's next for miramask

Our next steps include improving the val_accuracy of our ML model (which we would like to get to at least 98%), connecting the ML model to our UI, and eventually working with businesses to implement the project!

Built With

Share this project: