Skip to content
DemoFAQ

PDF Generation Template Editor

Open-source visual PDF generation engine with customizable templates and developer-friendly APIs.

View on GitHub

OpenVINO Acceleration Package ​

Overview ​

If your device uses an Intel processor, you can deploy the optional OpenVINO acceleration package to speed up AI recognition such as OCR. Once the package is deployed, the SDK automatically uses the OpenVINO backend when it loads the DocumentAI model, with no code changes and no extra settings.

Applicability ​

  • Operating system: Windows x64.
  • Processor: Intel CPU.
  • The package only affects the speed of AI inference (such as OCR recognition). It does not change any conversion API or conversion option.

Get the Package ​

Download the OpenVINO acceleration package that matches your SDK version from https://download.compdf.com/download. The package must match the SDK version; do not mix files from different versions.

Deployment ​

  1. Extract the downloaded package.
  2. Copy all files in the package to the directory that contains the SDK dynamic libraries, that is, the same directory as the onnxruntime runtime library.
  3. Load the model with LibraryManager::SetDocumentAIModel and run conversions as usual. The SDK uses the OpenVINO backend automatically.

Notice ​

  • If the package is missing, incomplete, or does not match the SDK version, the SDK automatically uses the default CPU backend. Conversion continues normally and no error is reported for this reason.
  • Deploy the whole package. Do not copy or replace only some of its files.
  • Deploy the package before the model is loaded. A model that is already loaded does not switch backends, so restart the program or reload the model after deployment.
  • The OpenVINO backend uses more memory. The actual speed-up depends on the processor model, document content, and thread settings, and not every scenario becomes faster.
  • To stop using acceleration, delete these files. The SDK then returns to the default CPU backend.