Swing

• 

  skeletal motion capture

Inspiration

To the beat of the rhythm of the night, dance until the morning light!   Unless you're like me and have no rhythm. Instead, we created a machine learning model to play music that matches your rhythm. So if you have no rhythm, do not worry, the machine will make you look good.

What it does

We used a tensorflow based recurrent neural network to classify dancing and play music that matches the dancer's beat per minute.

How we built it

Openpose for human pose estimation allowed us to motion capture ourselves dancing to various styles of music. We trained our own tensorflow based RNN in order to classify real-time dance moves and play music to match the beat.

Challenges we ran into

Turns out machine learning is pretty complex. We recorded over 24,000 data points in order to train our neural network and we should have gathered a lot more data.

Accomplishments that we're proud of

We trained and tuned a neural network at 4 AM.

Built With

• machine-learning
• python