Face Recognition Door with Deep Learning

Face Recognition Door with Deep Learning

Our project contains a camera, which recognizes familiar faces, and lets them in.


the code and the scripts can be found Here.


Edison instructions can be found Here.

Server instructions can be found Here.

Our work:
Image recognition is done with OpenFace, with additional small adjuesments we made.

Motion detection on edison is implemented by “Motion”.

All the rest of the scripts were implemented by us.

box components:
the model is not a part of the box

  1. Intel edison and arduino expension board
  2. Small light indicating the door being opened
  3. Webcam UVC conpatible
  4. Battery operated speaker
  5. USB hub
  6. USB sound card
  7. Jaw dropping beautiful box


Remove the use of keys, imporove home surveillance.


Images of welcomed guests are copied to the server (amazon EC2).

pose detection, alignment and training is performed.

the server has a docker container that waits for the images.

Once motion is detected by the camera, the magic begins! the camera starts taking pictures,

they are then uploaded to the server where image recognition is performed in the container,

using a deep learning algorithm (with OpenFace).

The server sending the answer- approved (with the person’s name) or not to the edison.

If the person is approved, the system will welcome him inside.


Ok, now seriously..

Possible future improvments:

  1. Improve overall speed.
  2. Create visitors log, including names and times.
  3. Send an email to the house owner upon a person’s arrivle.
  4. More user friendly DataBase creation.
  5. Improve the look of the box.
  6. Connect the system to an actual door!

Contact Us:
Loay Mualem: mualem.loay.1995@gmail.com
Mor Zamir: morzus90@gmail.com
Shiran Afergan: shiranaf@live.com

Close Close