Billiard Ball Mapper · Andreas Grill

1

Photo

A photo taken from directly above the table works best, but an angled shot is fine too — the next step corrects the perspective.

2

Mark the table corners

Drag the four dots onto the four corners of the playing surface (inside the cushions). This is what makes the map accurate regardless of the camera angle.

3

Map

Detection is a best guess — fix it up directly: drag a ball to nudge its position, tap a ball then the ✕ to remove it, or pick a color below and tap empty felt to add one. Striped balls are flagged but not numbered — this reads color and position, not printed numbers.

Add a ball

How it works

Runs entirely client-side with OpenCV.js (a WebAssembly build of OpenCV) — no photo is ever uploaded anywhere. The four corners you mark define a perspective transform that warps the table to a flat top-down rectangle; a Hough circle search on that warped image finds each ball, and its color is classified by sampling the pixels near its center against a reference pool-ball palette. It's a good first pass, not a certainty — the manual editing in step 3 exists specifically because lighting, glare, and odd angles can fool it.