Sesame

The smart lock to protect what you value the most, using face recognition as authentication.

Comment

Inspiration

In the world where technology is intricately embedded into our lives, security is an exciting area where internet devices can unlock the efficiency and potential of the Internet of Things.

What it does

Sesame is a smart lock that uses facial recognition in order to grant access. A picture is taken from the door and a call is made to a cloud service in order to authenticate the user. Once the user has been authenticated, the door lock opens and the user is free to enter the door.

How we built it

We used a variety of technologies to build this project. First a Raspberry Pi is connected to the internet and has a servo motor, a button and a camera connected to it. The Pi is running a python client which makes call to a Node.js app running on IBM Bluemix. The app handles requests to train and test image classifiers using the Visual Recognition Watson. We trained a classifier with 20 pictures of each of us and we tested the classifier to unseen data by taking a new picture through our system. To control the lock we connected a servo to the Raspberry Pi and we wrote C with the wiringPi library and PWM to control it. The lock only opens if we reach an accuracy of 70% or above. We determined this number after several tests. The servo moves the lock by using a 3d-printed adapter that connects the servo to the lock.

Challenges we ran into

We wanted to make our whole project on python, by using a library for the GPIO interface of the Pi and OpenCV for the facial recognition. However, we missed some OpenCV packages and we did not have the time to rebuild the library. Also the GPIO library on python was not working properly to control the servo motor. After encountering these issues, we moved the direction of our project to focus on building a Node.js app to handle authentication and the Visual Recognition service to handle the classification of users.

Accomplishments that we're proud of

What we are all proud of is that in just one weekend, we learned most of the skills required to finish our project. Ming learned 3D modeling and printing, and to program the GPIO interface on the Pi. Eddie learned the internet architecture and the process of creating a web app, from the client to the server. Atl learned how to use IBM technologies and to adapt to the unforeseen circumstances of the hackathon.

What's next for Sesame

The prototype we built could be improved upon by adding additional features that would make it more convenient to use. Adding a mobile application that could directly send the images from an individual’s phone to Bluemix would make it so that the user could train the visual recognition application from anywhere and at anytime. Additionally, we have plans to discard the button and replace it with a proximity sensor so that the camera is efficient and only activates when an individual is present in front of the door.

Share this project:

Updates