What it does ml4a-ofx is a collection of openFrameworks apps and Python scripts that let artists and creative coders use machine learning in real-time installations. The Python side handles heavy lifting like feature extraction and t-SNE; the C++ side handles frames and interactivity. Results shuttle back and forth via JSON or OSC.

The interesting bit The project doesn’t try to reimplement scikit-learn inside C++ — it admits Python is better for that, then wires the two worlds together. There’s even a dedicated OSC-modules folder for talking to Wekinator, a tool musicians already use to map gestures to sound parameters. It’s glue code, but thoughtful glue code.

Key highlights

• 15+ openFrameworks addons required, covering CV, audio, face tracking, darknet, and more
• Pre-built Python scripts for dataset prep and model analysis (VGG weights available separately)
• OSC bridge to Wekinator for performance and instrument control
• Each app has its own README with specific addon requirements
• setup.sh downloads example data so you’re not starting from zero

Caveats

• Windows and Linux setup docs are literally “[Feel free to add documentation here]” — Mac-only instructions for now
• The README’s own advice for build errors: “google the error (seriously!)”
• Addon dependency matrix looks intimidating; this is not a npm install experience

Verdict Worth a look if you’re already in the openFrameworks ecosystem and want to add ML without leaving it. Everyone else should probably start with a Python-first creative coding tool and revisit this when they need 60fps.