Inspiration
We took inspiration from security cam footage used to keep track of max capacity in indoor public spaces.
What it does
Our project detects whether or not people are wearing masks in camera footage using facial recognition software. The software is based on flask application and webdev, the model built with CNN image classification model.
How we built it
Using python flask application doc, and integrating html, js, css we built the ui and web interface. For the mask recognition we used keras to create a CNN model, we added sequential, conv and pooling layers then train and test models using 500 images of people with masks and 500 images of people without mask. These images were imported from Keras mask dataset. Then we saved this model and integrated into our code using the mediapipe engine.
Challenges we ran into
It wasn't easy to train our program to recognize key facial features with masks on. This involved understanding mediapipe api documents, and using flask for the first time.
Accomplishments that we're proud of
It was the first time for many of us to work with website design, and we are proud of what we accomplished in the end.
What we learned
We learned to train and work with neural networks, as well as the basics of front-end web design.
What's next for MaskOn
We hope to add additional features and improvemens to MaskOn, such as tracking volume of people, increasing accuracy and confidence,.
Built With
- css3
- html5
- javascript
- mediapipe
- python
Log in or sign up for Devpost to join the conversation.