30.09.2025
Smart Engines has enhanced its document recognition and verification technology with support for ID documents in Urdu and Persian. This advancement was made possible through the development of the world’s first specialized dataset for these languages—MIDV-UP—featuring more than 9,000 images of ID documents from Pakistan and Iran.

The dataset was unveiled at the International Conference on Document Analysis and Recognition (ICDAR). MIDV-UP addresses a long-standing gap in diverse, non-personalized training materials for documents in Urdu and Persian (Farsi). The dataset includes 1,000 unique samples representing several categories of identification documents: ID cards, driver’s licenses, and birth certificates from Iran, as well as national ID cards from Pakistan.
In total, MIDV-UP contains 9,000 fully annotated synthesized images that do not use personal data of real people. The dataset covers a broad range of document capture scenarios—from flatbed scans to photos and video sequences—featuring natural distortions such as shadows, glare, and perspective shifts. Smart Engines specialists used MIDV-UP to train a proprietary anti-fraud solution designed for advanced ID scanning and verification.
The company’s technology operates simultaneously in visible, ultraviolet, and infrared spectra, analyzing holograms and other security features and performing more than 600 checks. This approach allows the system to detect fraudulent documents of any complexity—from passports with altered or replaced photographs to sophisticated counterfeits and deepfakes.
Other news
30.09.2025Smart Engines Expands ID Document Scanning and Authentication Capabilities to Iran and Pakistan
16.09.2025Smart Engines launches GreenOCR® 2.0 with 10х higher accuracy and 20% faster document scanning
24.06.2025Smart Engines Unveils a Multimodal AI System for Document Forgery Detection
More news »
Send Request
Please fill out the form to get more information about the products,
pricing and trial SDK for Android, iOS, Linux, Windows.