Skip to content
Free public beta until December 4, 2026: every course is open, no card needed. After that, the first chapter of each course stays free and full access is a one-time $39 ($19 for up to 100 founding members chosen for helpful feedback). Your progress is kept. How the beta works

confidence score

A detector's own number, between 0 and 1, for how sure it is that a predicted box actually contains the object.

A confidence score is the probability, output by a detection model alongside every predicted bounding box, that the box both contains an object at all and is correctly classified. It is a learned estimate, not a guarantee -- thresholding it (keeping only boxes above some cutoff) is how a detector trades off missed objects against false alarms, and it is the number precision-recall curves (Section 4) are computed by sweeping.

Examples

  • A box drawn with confidence 0.92 is one the model is far more sure about than a box drawn with confidence 0.31; a pipeline might discard everything below 0.5.

See also

intersection over union (IoU)
Read the full section →
Graph settings

Calculator