8 Weeks have blazed past in the blink of an eye, and it’s now time to get cracking and focus on our prototypes! It is as exciting as it is intimidating. After seeing what the DIP students had presented, it was now up to us to match that standard.

Every freshie in SUTD learns Python, and to some very, very limited extent, machine learning. And thank god for it, because our theme decided to implement facial recognition in our lighting system.

What we wanted

After much discussion about user needs and how we could address them, we came up with 4 possible solutions:

1. Safety and Surveillance of children
2. Waste Management
3. Enhancing Family Time
4. Environmental Protection (Disturbance by kids, pets and insects)

We decided that it would be best to focus on both safety and enhancing family time as the hardware required for both were quite similar.

Facial Recognition (or not)

We managed to find an easy-to-use package that would allow us to run quick facial recognition scans, saving us the trouble of having to learn tensorflow and Facenet. And it even uses less than 10 lines of code!

Simply use

pip install tensorflow
pip install --no-dependencies easyfacenet

from easyfacenet.simple import facenet

images = ['images/image1.jpg', 'images/image2.jpg', 'images/image3.jpg']
aligned = facenet.align_face(images)
comparisons = facenet.compare(aligned)

print("Is image 1 and 2 similar? ", bool(comparisons[0][1]))
print("Is image 1 and 3 similar? ", bool(comparisons[0][2]))

(source)

And voila, easy comparisons! While this may be very simple (and a slightly lame use of Facenet), it was perfect for our project. With that, what we anticipated to be the most headache-inducing part of our project was more or less settled.

Design Sketches

Of course, no product is complete without an acceptable design. Our main problem now was to think of chassis to stuff our system into, and ensure that our design would be 3D-printable. However, it would have been counterproductive for us to directly model each design, only for it to be rejected. We settled on sketching what we had in mind, nevermind the our questionable sketching skills.

We intend to have 2 separate designs for the lights: one for the main entryway, and one to illuminate the courtyard itself. The entryway lights would do a simple check to see if someone is approaching the house – if there is, the entryway lights would light up to provide light, even for passer-bys. However, the in-courtyard lights would only turn on of the person is recognized as a resident, otherwise, it would send a notification to the owner.

Since it is more necessary for the inner lights to have cameras, we came up with various designs that could incorporate it in a relatively discreet way.