7-Stage Pipeline

How It Works

Your document passes through 7 verification stages in seconds. Watch the journey from upload to fraud detection.

or scroll down to explore each stage
STAGE 01
Image Upload
Upload front & back photos of the Greek ID card, plus an optional selfie for face matching.

Drag and drop or click to upload the front and back of a Greek National ID card. You can optionally include a selfie photo for identity verification. Supported formats include JPG, PNG, and WebP up to 10MB each.

STAGE 02
Quality & Preprocessing
Validates image quality, detects card edges, corrects rotation, and standardizes dimensions.

Each image is checked for blur, low resolution, and poor lighting. The system detects card edges, corrects rotation and perspective distortion, then standardizes to a fixed 856x540 resolution for consistent downstream analysis.

STAGE 03
Field Extraction
AI reads Greek and English text fields — name, date of birth, ID number, dates, nationality.
SurnamePAPAZIAN
NameANNA
DOB27.01.2003
ID No.A01877588

Advanced vision AI extracts all printed fields from both sides of the card: full name in Greek and Latin characters, date of birth, national ID number, issue and expiry dates, nationality, sex, and issuing authority. Each field is located by its known position on the standardized card.

STAGE 04
MRZ Validation
Machine-readable zone parsed character by character, every check digit independently verified.
I<GRCA018775882<<<<<<<<<<<<<<<
0301279F350615GRC<<<<<<<<<4
PAPAZIAN<<ANNA<<<<<<<<<<<<<<

The three-line MRZ on the back of the card is parsed following the ICAO 9303 TD1 standard. Every check digit is independently verified: document number, date of birth, expiry date, and the composite check digit. Any failed check is immediately flagged.

STAGE 05
Cross-Validation
6 independent checks comparing front-side fields against back-side MRZ data.
FRONT
BACK

Six independent cross-checks compare data extracted from the front of the card against the MRZ on the back: document number, surname, given names, date of birth, sex, and nationality. Any mismatch between front and back is a strong tampering indicator.

STAGE 06
Face Analysis
Portrait detection, anti-spoof verification, and selfie matching if provided.

The portrait photo on the ID is automatically detected and extracted. Anti-spoofing neural network analysis verifies it's a genuine photograph, not a printout or screen capture. If a selfie was provided, it's compared against the ID portrait for identity verification.

STAGE 07
Risk Scoring & Report
Fraud risk score 0-100 calculated, annotated images and full report generated.
0
CLEAN Document appears genuine

All findings from every stage are combined into a single fraud risk score from 0 to 100. Green (0-20) means Clean, amber (21-50) is Low Risk, orange (51-75) is Suspicious, and red (76-100) indicates Tampered. The final report includes annotated images highlighting every detection and finding.

All 7 stages complete. Document verified in under 10 seconds.