securAI: A New Approach to Home Security

My mom often says, "Be careful when you walk outside! They kidnap teens these days!" Using machine learning, we finally put her fears to rest, using a computer-vision powered, SMARTER security camera.

Comment

Inspiration

We were inspired the rising number of crimes in our neighborhoods, and the high price of modern security systems like Nest and ADT.

What it does

Our model takes a live feed from a Raspberry Pi and sends it to a flask server, which contacts the ML model to analyze the image and scan for unknown people. After the model makes its prediction, it stores its prediction in a Firebase database which is contacted by an app which sends notifications to alert users.

How we built it

For the model, we used Tensorflow. Our dataset was taken from Kaggle. With the Raspberry Pi, we used a camera to take photos, and those photos were converted into Base 64 before being sent to a Flask server. Firebase was used to store our testing data, which relied on a dataset containing images of popular celebrities.

Challenges we ran into

We had to edit our model many times over to successfully work in tandem with our camera. In addition, lots of fine-tuning was needed in order to improve accuracy. With the Raspberry Pi, encoding the images into Base 64 proved troublesome.

Accomplishments that we're proud of

What we learned

What's next for securAI: A New Approach to Home Security

Built With

Share this project:

Updates