FacePsy Mood Screening

On-device mood screening protects privacy by processing facial landmarks and action units locally in the browser, storing only anonymized encrypted vectors instead of raw images. It ensures facial features are extracted without transmitting visual data externally.

To perform a privacy-first depression screening, you scan your face in-browser to extract MediaPipe landmarks, 12 key action units, and head pose. The pipeline generates a mood risk score locally without storing or transmitting raw images.

Yes, facial action unit detection supports clinical mood tracking workflows by computing window-based features and hybrid universal-plus-personal models. It enables day-to-day mood trend monitoring for researchers and telewellness platforms while preserving patient privacy.

The best way to track depressive mood tendencies without storing raw images is using a privacy-by-design pipeline that extracts facial features on-device. It stores encrypted vectors locally and shares data only with explicit consent.

You do not need external libraries beyond the integrated client-side pipeline, which uses MediaPipe for landmark extraction and a dedicated AU CNN for action units. The system operates independently with no external dependencies.

A limitation of using facial features for PHQ-9 mood screening is that it detects mood tendencies rather than providing clinical diagnoses. It is designed as a non-intrusive screening support tool and should supplement, not replace, professional clinical assessment.