Other work — Machine Learning
ML5.js Pose Classifier
- Year
- 2025
- Context
- Creative Code 3 (Devine)
- Tools
- ml5.js, JavaScript, Canvas
A machine-learning experiment in the browser: webcam body tracking feeds three small classifiers that recognise a boxer's guard, stance and jab.
Case study
How it works
ml5.js tracks body keypoints from the webcam. You record labelled samples for three separate classifiers — guard (up, half, down), stance (orthodox, southpaw) and jab (correct, incorrect) — train them in the browser, and the page then predicts all three live. Trained models can be saved and loaded again.