OCR Extraction

Every field.
Structured.
In milliseconds.

Intelligent OCR that reads any identity document and returns clean, typed, confidence-scored data fields — no regex, no manual mapping, no post-processing. Just structured JSON, ready to use.

Request a demo

OCR Engine · Extracting fields

REPUBLIC OF INDIA · AADHAAR CARD

RISHI KUMAR SHARMA

XXXX XXXX 3847 · DOB: 14/03/1990

MALE · S/O RAJESH SHARMA · 42 MG ROAD, BANGALORE, KA 560001

"full_name""RISHI KUMAR SHARMA"

99.8%

"dob""1990-03-14"

99.9%

"gender""MALE"

100%

"uid_number""XXXX-XXXX-3847"

99.7%

"address""42 MG Road, Bangalore"

98.2%

"father_name""RAJESH SHARMA"

99.1%

"state""Karnataka"

99.5%

Fields extracted
All fields · 624ms · Aadhaar format

12 ✓

99.4%

Field accuracy

190+

Doc types

40+

Field types

<800ms

Response time

What We Extract

Every meaningful field. From any document format.

DigiVerify's OCR engine goes beyond simple text reading — it understands document structure, field semantics, and country-specific formats to return typed, labelled, validated data.

Personal Identity Fields

Core identity data extracted and typed precisely. Names returned in both full and split form. Dates normalised to ISO 8601 regardless of the source format on the document.

full_namefirst_namelast_namedobagegendernationalityplace_of_birth

Document Metadata Fields

All document-level data extracted and validated. Issue and expiry dates cross-checked against MRZ where available. Document numbers cleaned and normalised.

document_numberdocument_typeissuing_countryissuing_authorityissue_dateexpiry_dateis_expired

MRZ Zone Data

Full Machine Readable Zone parsing per ICAO 9303. Both MRZ lines extracted, checksums computed and validated, and all constituent fields decoded and returned individually.

mrz_line_1mrz_line_2mrz_validchecksum_passedmrz_document_numbermrz_dob

Address & Location Fields

Full address extraction where present on the document. Parsed into structured sub-fields — street, city, state, postal code — using country-specific address format knowledge.

address_fullstreetcitystatepostal_codecountry

Country-Specific ID Numbers

Dedicated extraction models for each country's unique identifier formats — validated against their official check-digit algorithms and format specifications.

aadhaar_idpan_numberssn_numbernhs_numberepic_numberghana_card_id

Biometric & Machine Data

The face photo is cropped, aligned, and returned as a URL ready for downstream face matching. QR codes and barcodes on documents are decoded and their payloads extracted.

face_image_urlqr_code_databarcode_datafingerprint_refsignature_present

How It Works

From raw image to structured data — in under 800ms.

The OCR pipeline runs four specialised stages in sequence, each optimised for accuracy over speed — and still returns results in under a second.

Image Pre-processing

The submitted document image is automatically enhanced before any text extraction begins. Perspective correction, de-skewing, glare removal, contrast normalisation, and resolution scaling are applied — maximising the quality of the input to the OCR models.

Document Type & Region Detection

The document type and issuing country are auto-detected from the image. The engine then identifies distinct regions — the personal data zone, photo zone, MRZ zone, barcode/QR zone, and address zone — and applies specialised extraction models per region.

Field Extraction & Parsing

Region-specific OCR models extract raw text from each zone. A semantic parsing layer then interprets the raw text into typed, labelled fields — splitting names, normalising date formats, validating ID number check digits, and decoding QR/barcode payloads.

Structured JSON Output

All fields are returned as a clean JSON object with typed values, per-field confidence scores, a list of unresolved or low-confidence fields, and cross-validation flags where MRZ data was used to verify visually-extracted fields.

Image Pre-processing

Perspective Fix

De-skew

Glare Remove

Contrast Norm.

Resolution Scale

Binarisation

Input quality: HIGH · DPI: 286 · Skew: 1.2° corrected

Document Type & Region Detection

Region Detection

IN Aadhaar Card — IndiaAUTO-DETECTED

📋Personal data zone

Found ✓

🖼️Photo zone

Found ✓

📱QR code zone

Found ✓

📍Address zone

Found ✓

Field Extraction & Parsing

Extracting & Parsing Fields

Name fields99.8%

Date of birth99.9%

UID number99.7%

Address98.2%

QR decode100%

Structured JSON Output

EXTRACTEDResponse: 624ms

{
  "document_type": "AADHAAR",
  "issuing_country": "IND",
  "full_name": "RISHI KUMAR SHARMA",
  "dob": "1990-03-14",
  "gender": "MALE",
  "uid_number": "XXXX-XXXX-3847",
  "father_name": "RAJESH SHARMA",
  "address_full": "42 MG Road...",
  "state": "Karnataka",
  "qr_decoded": true,
  "response_ms": 624
}

COUNTRY-SPECIFIC EXTRACTION

Trained on the exact fields each country's documents carry.

Generic OCR reads text. DigiVerify understands documents — it knows what Aadhaar fields look like vs a Ghana Card vs a PVC, and extracts accordingly.

India

UIDAI · IT Dept · ECI · MHA · ARTO

LIVE

Aadhaar Card

uid_numberfull_namedobgenderaddressfather_namestateqr_decoded

PAN Card

pan_numberfull_namefather_namedobsignature_present

Voter ID (EPIC)

epic_numberfull_namefather_namedobaddressconstituencypart_number

Passport

passport_numberfull_namedobexpiryplace_of_birthmrz_line_1mrz_line_2

Nigeria

NIMC · INEC · FRSC · CBN

LIVE

NIN Slip / NIN Card

nin_numberfirst_namedobgenderphonefingerprint_ref

PVC (Voter Card)

vin_numberfull_namedoblgastatepolling_unit

BVN Slip

bvn_numberfull_namedobphone

International Passport

passport_numberfull_namedobexpiryplace_of_issuemrz_valid

Ghana

NIA · EC Ghana · DVLA · SSNIT

LIVE

Ghana Card (NIA)

ghana_card_idfull_namedobgenderplace_of_birth

Voter ID (EC Ghana)

voter_idfull_namedobconstituencyregion

SSNIT Card

ssnit_numberfull_namedob

Ghanaian Passport

passport_numberfull_namedobexpirymrz_line_1mrz_line_2

Kenya

NIIMS · NTSA · Immigration Dept

LIVE

National ID (Huduma)

id_numberfull_namedobgenderdistrictdivision

Kenyan Passport

passport_numberfull_namedobexpirynationalitymrz_valid

Driver's Licence (NTSA)

licence_numberfull_namedobexpiryvehicle_classes

Alien ID Card

alien_idfull_namedobnationalityexpiry

Output Formats

Data in the shape your system needs.

DigiVerify returns extracted data in structured JSON with per-field confidence scores — typed, labelled, and cross-validated. Every response includes metadata on extraction quality and any low-confidence fields flagged for review.

Typed JSON

All fields returned with their correct data types — strings, ISO dates, booleans, enums. No string-to-date parsing in your codebase.

Confidence Scores

Every field carries a per-field confidence score (0–100). Low-confidence fields are flagged separately so your system can decide whether to auto-approve or queue for manual review.

Cross-validation Flags

Where MRZ data is present, visually-extracted fields are cross-validated against it. Mismatches are flagged with a discrepancy note — a strong signal of document tampering.

For Developers

Raw image in. Clean structured data out.

No template configuration. No regex rules. Pass an image and a country code — DigiVerify handles the rest and returns typed, labelled, confidence-scored fields.

Auto document type detection

Omit document_type and DigiVerify detects it automatically. Works reliably across all 190+ supported document types.

Bundled with Document Verification

OCR runs automatically as part of Document Verification — no separate call needed unless you want extraction-only without the full authenticity check.

QR & barcode decode included

Set decode_barcodes: true to automatically decode QR codes and barcodes present on documents like Aadhaar, PAN, and Ghana Card.

Front + back side support

Submit both sides of two-sided documents in a single call. DigiVerify merges the extracted fields from both sides into one unified response.

ocr-extract.js

// DigiVerify — OCR Extraction API
const result = await fetch('https://api.DigiVerify.com/v1/ocr/extract', {
  method: 'POST',
  headers: {
    'Authorization': `Bearer ${apiKey}`,
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({

    // Required
    document_image: documentBase64,
    country: 'IN', // IN | NG | GH | KE

    // Optional — auto-detected if omitted
    document_type: 'AADHAAR',

    // Back side for two-sided documents
    document_back: backBase64,

    // Return per-field confidence scores
    return_confidence: true,

    // Decode QR/barcodes present on doc
    decode_barcodes: true
  })
});

const {
  extracted_data, // typed fields object
  confidence_scores, // per-field scores
  low_confidence, // fields needing review
  cross_validated, // MRZ vs visual checks
  face_image_url // cropped photo URL
} = await result.json();

OCR is how you read the document. These are how you trust it.

Extracted data is only as good as the document it comes from. Pair OCR with authenticity checks to know the data is real.

Document Verification

OCR runs inside every Document Verification call — but Document Verification also authenticates the document, checks for tampering, and validates security features. Trust the source before you trust the data.

Learn more

Face Verification

Once the document is read and authenticated, the face photo extracted by OCR becomes the reference image for biometric matching — confirming the person matches the document data.

Learn more

Liveness Detection

Combine OCR data extraction with liveness detection to confirm the user submitting the document is physically present — not a replay or a fraudster using someone else's documents.

Learn more