Not The NSA

• 

  Facial Recognition Example!

Inspiration

We were inspired by machine learning software we encounter on a daily basis. We wanted to create a NSA thing that is a bit creepy in context but teaches us the fundamentals of API implementation and facial recognition software

What it does

Generator Function: Takes n amount of cropped frame shots of the subject's face via webcam. Each subject is uniquely ID'ed.

Trainer Function: Uploads each of the framed subjects' shots from the Generator and processes it to be recognized.

Recognition Function: Detects and recognizes each unique subject. Also sends a text message to the corresponding subject's phone to show them that we are watching them. Also detects if they are out of frame.

How we built it

We used a code built by a youtuber Anirban Kar which we manipulated for our use. We studied and did dozens of test cases to figure out what the best way is to train the software (more subjects and more photos of each subject, less subjects, more photos of each subject; more subjects, less photos; and so on and so forth).

We learned python fundamentals and machine learning principles in order to use a code and actually manipulate it to our use.

We also used the Telesign SMS API in order to send SMS messages to particular phone numbers.

Challenges we ran into

We tried doing the machine learning through Java and C++ before using Python. We spent hours on end doing trial and error. We also had a facial tracking component with an Arduino and Servo motor which worked, until ultimately for some reason the servo refused to track each subject. We also were going to do an authentication with Telesign which algorithmically worked, but the hardware-end faltered.

Ultimately, our teamwork prospered and held through with little sleep, but the biggest challenge was learning new technology and running into some hardware roadbumps.

Accomplishments that we're proud of

We are proud of doing machine learning, something we do not get to touch much. 2 of the members learned python; 1 of which is a Math major/CS minor who only took the CS minor recently. We are proud of the fact that we mentally pushed through each frustrating boundary and still have yielded a product that held the integrity we sought to hold: Use the Telesign API and have Facial Recognition (not just facial detection!). Ultimately, we are most proud that we learned a lot and are ready to conquer future hackathons to better ourselves and to share knowledge with other hackers.

What we learned

• Machine Learning
• Python
• Matrix mathematics
• Facial Recognition
• API implementation
• How to not cry when things don't work

What's next for Not The NSA

We are already preparing for the next Hackathons coming up, and are for sure going to finish this product over the summer. This will not be the last time we work together as a team and we will be sure to do our best at every event!

Built With

• machine-learning
• python
• telesign