I’ve been tinkering with different ways technology can shape digital and creative experiences:

• - Experimental three.js walk: an interactive “walkthrough résumé” built in 3D.
• - Noise reduction web app: browser-based audio cleaning experiment.
• - Text-to-speech prototype: exploring phonics and synthetic voice.
• - Interactive MIDI player: real-time browser music playback.
• - Headless Shopify integration: mock e-commerce build.
• - Artist portfolio: custom portfolio site for a performer.
• - Headless CMS site: CMS-driven publishing project.

These are prototypes and playgrounds: sometimes polished, sometimes half-finished, but all part of how I learn, experiment, and stay close to both creative practice and new technology.

Confusion Matrix

This matrix is encouraging. Most key predictions land cleanly on the diagonal — meaning the model is mostly getting things right. There’s some confusion between C and Eb, and a bit around Ab, but these are musically adjacent and share many pitches. That kind of blur is expected in real-world audio — not a red flag.

Importantly, we don’t see scattered misclassifications or random misfires. That suggests the model isn't overfitting or hallucinating. It's making sensible, musically interpretable errors.

Feature Importance

This bar chart shows the top 20 features driving XGBoost’s decisions. As expected, we see many strong contributions from individual CQT bins — likely centered around musically active registers. The presence of tonnetz_4 (a harmonic dimension) is a great sign that tonal structure is helping.

Even better: the new register-level summary features don’t dominate, but they make it into the top 20, confirming they’re useful without overwhelming the model.

Conclusion

These plots don’t look random, noisy, or overfit. They look like what I’d hope to see from a frequency- and harmony-aware key recognition system — grounded in both pitch content and musical logic.

I spent today improving my key classification project and got some pretty decent results! The model is trained to recognize the musical key of audio clips using features like chroma, tonnetz, and spectral contrast. I also made sure the top-3 predictions are returned with confidence scores, which helps when the model isn't completely sure.

Here's a snapshot of the confusion matrix to get an idea of how well the model performs across all keys:

And here's a bar chart showing which features the model thinks are most important. Pretty interesting to see that some chroma and contrast features matter more than others:

I'm still figuring things out but these visuals really help me understand how the model is making decisions. The pipeline also runs cleanly now and works from audio upload to prediction output. Will be adding more frontend tweaks and thinking about progression detection soon.