diff --git a/Sources/FocusCore/FocusScorer.swift b/Sources/FocusCore/FocusScorer.swift index a17265d..d279112 100644 --- a/Sources/FocusCore/FocusScorer.swift +++ b/Sources/FocusCore/FocusScorer.swift @@ -55,7 +55,7 @@ public struct ScorerParams: Equatable, Sendable { } } -/// Calibration knobs, ported from the ideas in e-candeloro/Driver-State-Detection (MIT). +/// Calibration knobs. /// ponytail: fixed constants tuned for a laptop webcam; expose in settings if users report false alerts. public enum Knobs { public static let calibrationSeconds = 10.0 diff --git a/docs/focus-scoring.md b/docs/focus-scoring.md index f2a1429..908f2f3 100644 --- a/docs/focus-scoring.md +++ b/docs/focus-scoring.md @@ -8,7 +8,7 @@ Each camera frame (about 10 per second) goes through Apple's Vision framework: ` ## Why not an end-to-end "focus %" model -End-to-end engagement models trained on DAiSEE reach about 60–73 % accuracy even in papers and generalize poorly. The established approach is pretrained face and landmark models plus transparent scoring, which is what [Driver-State-Detection](https://github.com/e-candeloro/Driver-State-Detection) (MIT) does and what FocusGuard follows. +End-to-end engagement models trained on DAiSEE reach about 60–73 % accuracy even in papers and generalize poorly. The established approach is pretrained face and landmark models plus transparent scoring. FocusGuard follows that approach. ## Calibration The first 10 s of each study phase record the user's baseline: neutral `yaw`/`pitch` (camera may sit low or to the side) and baseline EAR. All later checks use deltas from the baseline.