Release notes

Version 2.2.0 (2025-03-11)

API changes

  • Rectangle object changed from a struct to a class with utility methods

  • Added DocumentLibrary::getLoadedDocumentTemplate function

  • Added document_template_subset parameter to document detector

New features

  • New OCR model 3A, and improved post-processing

  • Blurriness detector to skip OCR on blurred images

  • Small improvements in Detector 2B algorithm

  • Harmonize detection scores for Detector2A and 2B. Warning: previous threshold might not be suitable anymore !

  • Improve handling of document templates, with possibility to unload with package name or document name

Bug fixes

  • Fixed Detector 2B bug on landscape images

  • Fixed detector crash when no template loaded

  • Windows Python Wrapper use OpenCL and runtime MT backend, fixing potential crash

  • Fixed bug on AlignDocument with low-scored detections

Version 2.1.2 (2024-11-29)

  • Fixed occasional crashes in DocumentReader

Version 2.1.1 (2024-11-15)

  • Fix AlignDocument not setting the Pixel Format

  • Catched misc OpenCV Exceptions and raise BackendError

  • [Python] Added Python 3.10 package on Windows and Linux

  • [Windows] Fixed cudart.dll dependency (which fixes [Python] ImportError: Missing DLL)

  • Internal tools update (codegen 3.38, image 3.13.1)

Version 2.1.0 (2024-07-12)

API changes

  • Added detailed options in ProcessingUnit enum for specific backends.

New features

  • Automatic model selection (among those loaded) in all modules, removing the need to call SetProcessingUnit or SetModel.

  • Improvements of OCR algorithms

  • New builds available for different new GPU backends:

    • Windows/Linux : default (OpenCL) & CUDA

    • Android : default (CPU Only) & GPU (OpenCL, Vulkan)

    • Apple : default (Metal)

  • All models now available for GPU use. Warning : This is experimental, some models might be unstable on some backends.

  • Internal tools update (codegen 3.36.4, image 3.8.1)

Bug fixes

  • Fixed minor bugs in DocumentReader

Version 2.0.1 (2023-12-22)

  • Fixed invalidDocumentTemplateFormat error when loading documents templates for detector 2B.

  • Fixed sporadic crash in DocumentReader

Version 2.0.0 (2023-12-21)

  • New unified multi-language documentation

  • Python support added

  • Swift support enhanced

New features

  • Model DocumentDetector1B renamed to 2A

  • New fastest detector model 2B for edge applications

  • New OCR model 2A, faster and better

  • Deprecated model DocumentDetector1A removed

  • Deprecated model MRZReader1A removed

  • New organisation and naming for Document Templates

API Changes

  • License module renamed DocumentLicense.

  • DocumentLibrary license methods removed.

  • DocumentLibrary.getVersionEx() method added.

  • Image renamed DocumentImage.

  • ReferenceDocument renamed DocumentTemplate.

  • DocumentPose renamed DocumentOrientation.

  • DocumentDetector.DetectAndClassifyDocument() renamed DetectDocument().

  • DocumentDetector.DetectDocument() renamed DetectDocumentByName().

  • DocumentInfo class added.

  • DocumentCategory enumeration added.

  • DocumentFormat enumeration added.

  • Internal updates license v8.6.1, codegen v3.24.0 and image v3.7.1.

Version 1.1.2 (2023-08-24)

  • Added pattern correction for OCR

  • Added field rotation [+90°, -90°, 180°] for OCR

  • Internal tools update (codegen 3.21.1)

Version 1.1.1 (2023-08-02)

  • Fixed bugs in DocumentDetector

  • Fixed bug when loading concatenated document template

  • Updated new Python wrapper

  • Fixed TextField bounds not implemented

Version 1.1.0 (2023-06-19)

  • Added Optical Character Recognition (OCR) feature

    • Added new model OCR v1A for latin alphabet and digits

    • Added module DocumentReader to read text fields on a document image

Version 1.0.2 (2023-05-11)

  • Internal tools update (codegen 3.13.1)

Version 1.0.1 (2023-01-27)

  • Fixed MRZ validity check for TD1 type with long document numbers

Version 1.0.0 (2022-12-21)

  • Initial release.