A desktop game where you dance salsa in front of your webcam, and computer vision judges your moves!

Real-time pose estimation (YOLO) feeds a LSTM move classifier that recognizes your dance moves (front_and_back, idle, side_step, spin) live from webcam video. Dance well and you'll earn your salsa license from Señorita Cabí and Señor Empanada, up on the slopes of Monserrate, Bogotá.

1. Grab the latest .dmg from Releases.
2. Open the .dmg and drag Macondo into your Applications folder.
3. The app is ad-hoc signed, so Gatekeeper will block the first launch as "unidentified developer." Right-click (or Control-click) the app in Applications and choose Open
4. Grant camera access when prompted

Only an Apple Silicon (arm64) build is currently published; there's no Windows/Linux release yet, see Frontend setup below to run from source instead.

• Dance clips are recorded per-move with backend/dataset_recorder.py into backend/dataset//e>/.
• backend/train_lstm.py trains the LSTM classifier on that dataset and writes a checkpoint to backend/model/lstm_move_classifier.pt.
• backend/test_lstm_model.py opens a live webcam window with prediction bars.

• Python 3.12+ and a virtualenv at .venv/ in the repo root
• Node.js + npm
• A webcam
• A Supabase project (for the leaderboard)

Apply supabase/schema.sql to your Supabase project to set up the leaderboard table and RPC.

python -m venv .venv

source .venv/bin/activate    # Windows: .venv\Scripts\activate
pip install -r requirements.txtt

cd frontend
npm install
cp ../.env.sample .env     # fill in your Supabase URL + publishable key

Run everything together as the Electron app (starts the Python backend automatically and opens the game window):

npm run electron:dev

Thank u so much to Tati and Dhamari for helping train our model