Remember when you lived in a dorm? I thought of this as a brilliant way to spice up dorm life - someone knocks on your door and boom, JOHN CENA

Everyone gets their own theme song - you'll never have to guess who's covering up your peephole, you can certainly tell the difference between Nyan Cat and Kazoo Kid.

What it does

It uses openCV and Haar Cascade to recognize faces, then links to Mopidy, a music playback daemon, on my Raspberry Pi. User profiles are stored locally via JSON, and the app automatically searches for the right audio file to play.

How I built it

I put the peeps in the chili pot

I wanted to do some ML, and I thought this hack would be hilarious, fun, and useful to me whenever this quarantine ends, so I put everything together in one .py file.

Challenges I ran into

  1. The Pi kept running out of graphic memory.
  2. The Pi took years to train.
  3. My initial code would recognize the face and then produce the wrong output.
  4. The Mopidy integration would freeze the video output for overly long periods of time - I managed to limit it to just the crossfade duration.

Accomplishments that I'm proud of

This was my first ever Raspberry Pi experience, and I'm pleased to say it worked beyond my expectations. The little thing just kept chugging along no matter how many matrix multiplications I made it do. Also, setting up the Mopidy server was pretty rough, and I have yet to get a stable Spotify connection, but it works perfectly with MP3 files.

What I learned

OpenCV makes life easier than I thought,

What's next for Hype Man

Full Spotify integration! SSH support and app integration, so you can start the Pi from your mobile phone

Try It out



machine-learning, mopidy, opencv, picam, pygame, python, raspberry-pi, spotify

Devpost Software Identifier