What is new in FlexiCapture Engine 10 Windows

ABBYY FlexiCapture Engine 10 is the newest version of ABBYY's Data Capture SDK, designed to power a new generation of business applications. Its new and enhanced features offer the fastest way for developers to integrate sophisticated data and document capture technologies into server, desktop, mobile and cloud solutions. A highlight of key new features includes:

  • New Auto-Learning Technology
  • New Auto-Classification Technology
  • Enhanced Image Pre-processing
  • Adjusted Licensing Scheme

Auto-Classification Technology and Auto-Learning

ABBYY FlexiCapture Engine 10 makes it easy to start working with sophisticated classification and capture technology. Now you can train document classifiers and create simple document layout descriptions in just minutes.

Automatic Document Classification

The new FlexiLayout Studio 10 classifiers can be easily created and trained on the batch of pages which represent different classes. The new classification technology can be set up via FlexiLayout Studio or via API.
For this purpose the image-based and content-based analysis may be used.

To train the Classes automatically it is necessary to select the “Auto” classification mode. It is possible to use different profiles for training:

  • The Raster profile compares image appearance using special algorithms.
    • The Raster (fast)profile is similar to Raster with the difference that it doesn’t take page size, page width to height ratio, etc. into account.
  • The Raster with Titles profile.
    • In addition to the Raster profile, it compares images and text of large size recognized on images.
  • The Raster with Titles and Image Objects profile additionally uses different objects (separators, barcodes, etc) for comparison.
    • The Raster with Image Objects profile uses Raster + such objects as separators and barcodes for comparison.

Automatic Learning Technology

The new Auto-Learning API allows developers to create a simple point-and- click user interface, so they can extend applications and automatically create data extraction rules. The application will “learn” how to extract data from simple, one page documents, without the need to use FlexiLayout Studio.

Here a screen cast that shows the general principle

Note: the latest code sample that comes with FlexiCapture Engine 10 has already changed - a new video is in preparation ;-)

A step by step description can be found here: Creation Document Definitions by Training

Enhanced Core Technology

New Image Pre-processing Tools

  • Remove security paper guilloche background or suppress colour objects in official documents.
  • Remove garbage from an image.
  • Smooth images textures.
  • Create different filters based on frequency domain transformation of an image to extract maximum information from difficult to read documents.

Better recognition results on difficult images, for example the guilloche background of an ID card

New features of Camera OCR

Camera-based document acquisition enables new and emerging applications such as mobile capture and remote capture with server-based processing. The new and enhanced features are designed to enhance text recognition and data extraction from documents captured using digital cameras. The new set of document photo adjustment features include:

  • Remove noise from colour photos.
  • Remove motion blur from photos.
  • Remove geometrical distortions from an image.

New Binarization

New binarization technology provides better text retention and prevents information loss even in difficult cases. Enhanced binarization quality and retention of valuable information for better OCR accuracy. Read more...

Development Improvements

Recognition Variants

FlexiCapture Engine 10 introduces a native API to access to different OCR hypotheses for character and word recognition. The information can be used in external voting algorithms so that data accuracy can be improved without manual verification.

Enhanced Java Support

Version 10 makes it easier for developers /enterprises to extend their JAVA applications (on Windows) with data capture functionalities. FlexiCapture Engine 10 includes a new Java Native Interface (JNI).

New pre-compiled Code Samples

  • Image Pre-processing Sample
  • Auto Template Creation with out FlexiLayout Studio

FlexiLayout Technology Improvements

FlexiLayout Studio Improvements

  • Training a FlexiLayout on sample pages
    • A FlexiLayout can be created automatically based on fields and static elements marked on several pages.
    • This feature of FlexiLayout Studio is actually used when you “train” documents via the API and your custom GUI
  • Training an auto-classifier
    • The Decision Tree can now train itself on a set of pages with specified reference classes. A few pages of each class are included in the training set, while other pages can be used to test the trained auto-classifier. For training an auto-classifier you can use pre-defined profiles based on different criteria.
    • The same functionality is also available via the Document Definition Training API
  • Classification log
  • FlexiLayout Language
    • New functions for data type conversion
    • New functions of access to a recordset from a database have been added
    • The Keyword method that retrieves a keyword by its index in the string has been added for the Static Text element
    • Methods for calculating the intersection area of the objects and a speciried rectangle have been added for the Page type

Further Improvements

  • Trainable properties of elements
  • Trainable relations
  • Trainable offset
  • FlexiLayout AutoCreation

Further Improvements

  • CJK search mode
    • The new search mode improves finding text lines on Chinese, Japanese and Korean languages.
  • Extended 2D barcode support:
    • QR Code added

Adjusted Licensing

The pricing and licensing scheme for FlexiCapture Engine 10 has been revised and updated.

  • FlexiCapture Engine Runtimes are now delivered as network licences. The licensed page volume or the licensed number of CPU Cores can be rolled out in the LAN. Daily peak volume can easily be distributed within a network. Virtual machines running on blade servers can easily be used for up scaling.
  • ABBYY Europe changed the time period for the page counter from “pages per month” to “pages per year”. The volumes were adjusted to align with ABBYYs FlexiCapture Server’s licensing. The new default delivery of the page volume makes it easier to deal with peak volumes.
  • FineReader Engine 10 API is now included 1). FlexiCapture Engine 10 allows to apply FlexiLayouts for separation and data extraction, and additionally to call the regular layout analysis provided by FineReader Engine, including all export formats, like text only, Office formats or Full page XML.
    The new licensing scheme is perfectly suited for existing FineReader Engine developers who want to extend their existing applications with intelligent separation and data extraction features.



Back to FlexiCapture Engine 10 Overview FlexiCapture Engine 10 Tech Specifications

1) included in the new licensing model from ABBYY Europe; it is still an add-on in other ABBYY sales territories.
This website uses cookies which enable you to see pages or use other functions of our websites. You can turn off such cookies in your browser’s settings. If you continue to use these pages, you consent to the use of cookies.