CAPA SDK Documentation

Craniofacial Analysis & Prediction Architecture

Welcome to the CAPA SDK documentation. This SDK provides advanced craniofacial analysis capabilities based on 15+ peer-reviewed scientific papers.

Getting Started - Installation and quick start guide
API Reference - Complete API documentation
Financial Verticals - Phase 1 biometric pipelines (KYC, continuous auth, anti-deepfake)
Scientific Foundation - Research papers and methodology
Examples - Usage examples and tutorials
Configuration - Configuration options

Overview

CAPA SDK provides the following analysis capabilities:

Core Features (craniofacial analysis)

Feature	Description
WD Analysis	Bizygomatic width measurement and social orientation classification
Forehead Analysis	Frontal inclination angle and impulsiveness level assessment
Morphology Analysis	Face shape classification and facial proportion analysis
Neoclassical Canons	Classical facial proportion validation
Multi-Angle Analysis	Combined analysis from multiple image angles

Financial Verticals (Phase 1 — biometric pipelines)

White-box, audit-friendly pipelines built on top of the morphology and biometric primitives. All decisions carry decision_reason, provenance and a structured audit_trail.

Module	Class	Purpose
`capa.biometric`	`FaceMatcher`	ArcFace (InsightFace `buffalo_l`) face verification, cosine threshold 0.40
`capa.biometric`	`LivenessAnalyzer`	Passive liveness fusing 3D structure + LBP texture + radial FFT + temporal coherence
`capa.kyc`	`KYCPipeline`	Selfie + ID document → identity match, liveness, MRZ parsing, audit trail
`capa.continuous_auth`	`ContinuousAuthSystem`	Per-user baseline + drift detection (z-score over enrolment self-similarity)
`capa.deepfake`	`DeepfakeDetector`	Multi-frame fusion: identity consistency + mesh stability + liveness + spectral + blink dynamics

Quick Example

from capa import CoreAnalyzer

# Initialize
analyzer = CoreAnalyzer()

# Analyze
result = analyzer.analyze_image("photo.jpg")

# Access results
print(f"WD Value: {result.wd_result.wd_value}")
print(f"Face Shape: {result.morphology_result.shape_classification.primary_shape.value}")

# Cleanup
analyzer.shutdown()

Architecture

capa/
├── analyzers/          # Analysis orchestrators
│   ├── CoreAnalyzer
│   ├── MultiAngleAnalyzer
│   └── ResultsIntegrator
│
├── modules/            # Scientific analysis modules
│   ├── WDAnalyzer
│   ├── ForeheadAnalyzer
│   ├── MorphologyAnalyzer
│   └── NeoclassicalCanonsAnalyzer
│
├── biometric/          # Phase 1 — biometric primitives
│   ├── FaceMatcher           (ArcFace verification)
│   └── LivenessAnalyzer      (passive liveness, multi-modal)
│
├── kyc/                # Phase 1.A — KYC pipeline
│   ├── KYCPipeline
│   └── DocumentProcessor     (face crop + MRZ extraction)
│
├── continuous_auth/    # Phase 1.B — Continuous authentication
│   └── ContinuousAuthSystem  (baseline + drift detection)
│
├── deepfake/           # Phase 1.C — Anti-deepfake
│   └── DeepfakeDetector      (multi-frame video analysis)
│
└── _internal/          # Internal support systems
    ├── IntelligentLandmarkSystem
    ├── AdaptiveQualitySystem
    └── ContinuousImprovementSystem

Requirements

Python 3.9+
OpenCV
dlib
MediaPipe
face-recognition
scikit-learn
scipy
numpy
insightface + onnxruntime (Phase 1 — biometric)
passporteye + tesseract (Phase 1.A — KYC MRZ)

License

Dual License (Non-Commercial Free / Commercial Paid)

Use Type	License	Cost
Personal projects	Non-Commercial License	Free
Academic research	Non-Commercial License	Free
Educational use	Non-Commercial License	Free
Non-profit internal use	Non-Commercial License	Free
Commercial use	Commercial License Required	Contact us

Patent Notice

This software implements methods covered by US Patent 10,885,309: "System and method for evaluating personality using anthropometric measurement of a person's face"

For commercial licensing inquiries, contact: yarelleyva2@gmail.com

See the full LICENSE file for complete terms.

Support

For issues and questions, please visit our GitHub repository.