Home▸Machine Learning & Neural Networks▸3D Confusion Matrix Threshold Lab

3D Confusion Matrix Threshold Lab

A 3D point cloud of classifier scores, colored live by TP/FP/FN/TN as you drag a decision-threshold plane through it, with a 3D confusion-matrix bar cluster and real precision, recall, F1 and accuracy readouts.

Machine Learning & Neural Networks3DEasy60 FPS📱 Mobile-adapted⇄ 2D version

Every sample in this scene is a point positioned by its classifier score, and every point knows its own true class. Drag the translucent decision plane through the field and each point recolors the instant the plane crosses it — green for a true positive, red for a false positive, amber for a false negative, gray-blue for a true negative — while a cluster of four bars behind the field renders the confusion matrix itself in 3D, height tied directly to a live recount. Precision, recall, F1 and accuracy in the sidebar come straight from those same four counts, so orbiting the scene and watching individual points flip category is the same operation as reading the numbers — just made physical.

⚙ Under the hood

Drag a decision plane through a 3D point cloud of classifier scores and watch each sample recolor live as a true/false positive/negative, while a 3D bar cluster renders the confusion matrix and precision, recall, F1 and accuracy update from real counts.

confusion-matrixprecision-recallclassification-metrics

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install