Traditional music recommendation depends on explicit choices and does not respond to a person’s immediate emotional context.
Inside the experience
What someone can actually do
- Real-time facial-expression pipeline
- Emotion-specific music datasets
- KMeans recommendation step
- Published research connection
How the idea took shape
The prototype uses OpenCV and a FER-2013-trained model to classify facial expression, then combines clustering and Spotify data for mood-aligned recommendations.
“Context can make a recommendation feel more responsive, but sensing emotion also raises questions about ambiguity, consent, and interpretation.”
Explore how it was built +
I built the facial-expression pipeline, organized emotion-specific music data, and connected classification with clustering and Spotify-based recommendations.
- Python
- OpenCV
- Keras
- scikit-learn
- Spotify API
- PHP
Where it could go next