ud-synthetic/belgian-passports

Disclaimer: All passport images and associated data in this dataset are synthetically generated and do not correspond to real individuals. Any names, numbers, or personal details are fictional and used solely for research and development purposes.

Introduction - Belgium

The Synthetic Belgium Passports Dataset brings together more than 1,000 AI-generated passport images built for training OCR and computer vision systems on identity documents. Every record is fully synthetic, so the collection contains no real personal data or biometric records — a privacy-compliant foundation for building identity verification, KYC, and fraud-detection pipelines. - Get the data

The dataset can be expanded on demand — additional images and metadata can be produced to match your specifications, typically within one week.

Each image is taken on either a clean white background or a mix of realistic settings — desks, walls, and other everyday surfaces — helping models generalise more reliably to production-grade conditions.

Coverage spans 50+ countries (including Netherlands, Luxembourg, France, Germany, Austria, and more). Submit a request via the website to learn more.

Each image is supplied with rich structured metadata describing the document's personal fields — passport number, full name, signature, date of birth, sex, place of birth, issuing authority, nationality, document type, and the machine-readable zone (MRZ) — together with technical attributes such as resolution and category.

---

Dataset general info

Characteristic	Data
Description	Synthetic passport images with detailed metadata for ML model training in PII extraction
Data types	Image + structured metadata
Tasks	OCR, Computer Vision
Total number of files	1000+ (scalable on request)
Labeling	Passport Number, Passport Number (split), Surname, Given Name, Signature, Date of Birth, Date of Issue, Date of Expiry, Sex, Place of Birth, Issuing Authority, Nationality, Nationality Code, Document Type, MRZ
Gender	Male, Female
Backgrounds	White and varied (desk, wall, and other surfaces)
Countries	50+ available (Netherlands, Luxembourg, France, Germany, Austria, and more — on request)
Image format	JPG
Data generation	AI-generated
Source of images	AI-generated

The dataset comprises 1,000+ synthetic passport images, each tied to a complete identity-style record and structured annotations. Additional samples can be produced on request within one week, and the imagery covers both clean white scenes and varied background environments.

Metadata fields include:

Field	Example
Passport number	EM1234567
Passport number (split)	E M 1 2 3 4 5 6 7
Surname and given name	PEETERS LUC
Signature	Luc Peeters
Date of birth	22 SEP 1985
Date of issue	14 JUN 2022
Date of expiry	14 JUN 2029
Sex	M
Place of birth	BRUSSELS
Issuing authority	MIN. AFFAIRES ETRANGERES
Nationality	BELGIUM
Nationality code	BEL
Document type	P
Machine-readable zone (MRZ)	P<BELPEETERS<<LUC<<<<<<<<<<<<<<<<<<<<<<<<<<<EM12345672BEL8509223M2906141<<<<<<<<<<<<<<00

---

Use cases - Belgium

Training document verification systems

Banks, fintech firms, and border-control units can apply this dataset to train models that verify identity documents across a wide variety of real-world conditions. Structured metadata supports accurate field extraction and validation, while a mix of backgrounds reinforces robustness once models are deployed.

Building fraud detection pipelines

Compliance and security teams can rely on these synthetic samples to compose high-quality training corpora without involving real travel documents. Full passport field coverage and complete MRZ strings allow both standard records and anomalous patterns to be modelled and tested with confidence.

---

FAQ

Is this real-world or synthetic data?

All images are AI-generated and contain no biometric data or personal information tied to real individuals.

Can I request a custom dataset size?

Yes — the dataset is scalable, and additional samples can be generated based on your requirements within one week.

Can I request country-specific data?

Yes — support for 50+ countries is available. Please submit a request to get detailed coverage and samples.

Can I request a sample before purchasing?

Yes — free samples are available so you can evaluate image quality, metadata structure, and variation coverage before committing.

How is the dataset delivered?

After purchase, the dataset is delivered via secure AWS cloud infrastructure compliant with ISO 27001 and ISO 27701.

🌐 UniData - your trusted data partner. Unique, accurate, thoroughly collected and annotated data designed to fuel your AI/ML success.