Aspose.OCR for .NET 新版本發表

Released: Mar 31, 2026

V26.3 中的更新

修補程式

  • Performance and memory consumption issue: add 'EnableCpuMemArena' and 'EnableMemoryPattern' to configure ONNX Runtime.

Released: Feb 27, 2026

V26.2 中的更新

特性

  • Internal improvements and stability refinements.

Released: Jan 28, 2026

V26.1 中的更新

特性

  • Improved region detection accuracy.
  • Keyword and numeric search - Allowed searching by keywords as well as numeric values.
  • Added RecognizeTables method in the API.

修補程式

  • Fixed bug with resource download (add check sum for models size).

Released: Dec 27, 2025

V25.12 中的更新

特性

  • Integrated the new LLM AI library into the project.
  • Removed all obsolete fields and functions.

Released: Nov 18, 2025

V25.11 中的更新

特性

  • Added CSV output format support for table detection (in OcrOutput class).
  • Added support for OCRTable recognition result format.
  • Integrated formula detection and recognition functionality.

Released: Sep 25, 2025

V25.9.1 中的更新

特性

  • Table recognition has been improved.

修補程式

  • Fixed issue with OCR Accuracy using Aspose.OCR on PDF.

Released: Sep 12, 2025

V25.9 中的更新

特性

  • Improved markdown output format with table detection.

修補程式

  • HOCR export to MemoryStream and incompatibility with Aspose.PDF conversion.

Released: Aug 29, 2025

V25.8 中的更新

特性

  • Integrated LLM into table recognition.
  • Improved Excel output with LLMs table recognition.

修補程式

  • Fixed bug with multi languages and multi threading.

Released: Jul 13, 2025

V25.7 中的更新

特性

  • The new AsposeAI class provides the bridge between LLM models and the OCR engine. The AsposeAI class integrates external AI models (e.g. from Hugging Face) into OCR recognition results for spell-check correction and intelligent post-processing.

Released: Jun 13, 2025

V25.6 中的更新

特性

  • Added a debug mode to the API to allow users to view detection area results on the image.
  • Added 'confidence' setting for text lines in RecognitionResult.
  • Updated ONNX Runtime to version 1.22.0.

修補程式

  • Fixed hOCR output formatting error.