Calibrate a camera frame

Calibrate a camera's view of a court, table or screen once from four corners, so every later point is measured in that frame

Give min_confidence (the floor an automatic result must reach) and detected ({ corners: four [x, y] in the order top-left, top-right, bottom-right, bottom-left, confidence }) as a corner detector gave it, or null; manual_corners (four [x, y] from a guided manual adjustment) are used when the automatic result is missing, too unsure or unusable. points ([[x, y]]) are mapped into the frame. Answers state (calibrated or needs_manual), source (auto or manual), the confidence seen, h (the projective map to the unit square) and failures: each failed automatic attempt with its reason, for the caller to keep as training data. Corners that are repeated or in a line are refused in words. It runs no detector and keeps nothing.

What it asks for

Nothing: your AI starts it with your words.

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.

Say to your AI: “Use Trillion to calibrate a camera's view of a court, table or screen once from four corners, so every later point is measured in that frame”