29.11.2019 г.

Smart Engines document capture system has mastered handwriting recognition

Smart Engines has introduced a new version of its proprietary AI system Smart IDreader 3.3. The major feature of version 3.3 is the ability to capture and extract handwritten fields of the identity documents. Apart from that, among the improved capabilities is the recognition accuracy, which refers to over 1,100 document types from all over the world that the system supports.

Smart IDreader is a software designed to automatically extract data from various identity documents in mobile applications, web-solutions and front-office systems. The developer of the system is a company Smart Engines which specializes in software development based on AI and machine vision.

The cutting-edge scientific advances applied in the development of the new version, fundamentally optimize the Optical Character Recognition (OCR) process. Smart Engines developers refer to their technologies as Green OCR; the name derives from the radically new approach taken by Smart Engines, to the implementation of recognition mechanisms. The approach per se is aimed at minimizing energy consumption during the neural network training process, as well as reducing the power consumption of devices on which the calculations associated with data capture are performed. Thus, the newly released software allows performing complex recognition tasks in a power productive way with the reduced number of the neural networks used.

Handwritten fields present a challenging task for data capture systems in the document analysis field. The transition from extraction of printed characters and texts from images (photographs and scans) to the processing of hand-written characters is a serious result in the development of AI.

The updated version is distinguished by facilitated integration and improved quality characteristics, namely: recognition accuracy, the ability to detect and choose the best shot in the video stream, and updated algorithm which is responsible for checking the fields of the passport or VISA, for further cross verification with its MRZ field.
Furthermore, Version 3.3.0 introduces a new “ID Shot” module – a compact SDK for capturing ID images on mobile devices in a video stream. Now, users can only search and capture a document image on a mobile device, while the entire computationally intensive recognition process will be performed on the server.

Another added feature is the ability to recognize more than 50 types of identity documents such as driver’s licenses, passports of different countries, including previously unsupported ID-cards of Argentina, Jordan and Montenegro. The system has updated scripts for automatic assembly of the modules, which allow integration of server SDK with PHP, Python and UiPath Robotic Process Automation.
Smart IDReader supports SPARC, MIPS, ARM, x86 hardware platforms and is compatible with the iOS, Android, Linux, Windows, macOS, Solaris operating systems.

Among the users of the technology Beeline, Megafon, MTS,Gazprombank, Alfa-Bank, Post Bank, Tinkoff, Alfastrakhovanie, Renaissance Insurance Group, Aeroexpress, and many more.

You can test the capabilities of the new version by installing the free Smart IDReader demo application from Google play or the App Store.