Craniofacial Analysis SDK with biometric demos
A Python SDK for craniofacial analysis built on 15+ peer-reviewed papers, with interactive demos that show how it can be applied to face-biometric use cases such as KYC, continuous authentication and anti-deepfake. Powered by ArcFace, MediaPipe and dlib.
15+
Scientific Papers
7
Craniofacial Modules
3
Biometric Pipelines
4
Interactive Demos
Core SDK
Craniofacial Analysis
Seven science-backed modules for facial measurement and interpretation
WD Analysis
Bizygomatic width (fWHR) measurement and social personality trait correlation based on peer-reviewed research.
Forehead Analysis
Frontal inclination measurement and impulsiveness assessment with neuroscience correlations.
Morphology Analysis
Face shape classification (10 types) and comprehensive facial proportions analysis.
Neoclassical Canons
Analysis of 8 classical facial proportions based on Farkas principles with ethnic group support.
Ear Detection
Automatic ear localization and geometric measurement to support profile and symmetry analysis.
Profile Analysis
Lateral-view assessment of facial angles and convexity from optional profile photographs.
Identity Verification
InsightFace ArcFace 1:1 matching with cosine-similarity threshold and quality gating.
Face Biometrics
Biometric Capabilities
Face-biometric building blocks combined into the demo pipelines
KYC Onboarding
End-to-end pipeline: ID document OCR (passporteye), face vs. document match (ArcFace) and passive liveness scoring.
Continuous Authentication
Drift detection over rolling embedding windows with configurable thresholds and per-user enrollment baselines.
Anti-Deepfake
Five-signal weighted detector (texture, blending, frequency, blink, optical-flow) with margin-based decision policy.
Passive Liveness
Single-image liveness scoring to flag spoofing attempts such as printed photos and screen replays.
Configuration
Analysis Modes
Six execution modes to balance speed, depth and scientific rigor
Quick analysis for preview — basic modules only
Balanced speed and precision — recommended default
Deep analysis with all modules and 3D reconstruction
Maximum accuracy — measurable 2D observables only
Peer-reviewed correlations with explicit disclaimers
Optimized for low-latency real-time processing
CAPA v1.1.0