Bluey

Aggregate metrics

Based on 7 episode(s). FFC: 0.192

Unlike black-box tools, CMAT measures its own error rate against human coding. That validation is currently a small, single-coder pilot, so treat the figures as indicative, not settled. Detection accuracy (preliminary): transition-BOUNDARY detection across all coded transition types — i.e. whether a transition was found there, not whether it was labelled correctly — agreed with human coding at F1 0.85 (±2s match, 2 comparison run(s), single coder; larger sample and inter-rater reliability in progress). This is the figure the pacing RATE depends on; type classification is scored separately and is lower. Accuracy is content-dependent. Dissolve/scene-change detection experimental. Colour, motion and audio are deterministic measurements; FLASHING is unvalidated and is not a safety assessment. See Methodology.

Episodes