Property OcrEngine

Namespace

LMKit.Extraction

Assembly

LM-Kit.NET.dll

OcrEngine

Gets or sets an optional OcrEngine used to perform traditional OCR on raster content.

public OcrEngine OcrEngine { get; set; }

Property Value

OcrEngine

Examples

using LMKit.Extraction;
using LMKit.Extraction.Ocr;

var te = new TextExtraction(LM.LoadFromModelID("lmkit-tasks:4b-preview"));

// Any concrete implementation of OcrEngine can be used here
OcrEngine ocr = new MyOcrEngine(); 
te.OcrEngine = ocr;

te.Elements = new List<TextExtractionElement>
{
    new TextExtractionElement("Total", ElementType.Decimal)
};

te.SetContent(new Attachment("invoice_scan.png"));
var result = te.Parse();
Console.WriteLine(result.GetValue<decimal>("Total"));

Remarks

This property is optional. If no OCR engine is provided (i.e., left null), image-based text recognition will rely solely on the language model's vision capabilities.

When set, OCR may be invoked automatically during extraction depending on the PreferredInferenceModality and the structure of the input:

• If PreferredInferenceModality is Vision or Multimodal, the engine will attempt to extract text from any raster content before invoking the language model.
• When the input includes both text and images, providing an OCR engine can substantially improve recognition accuracy by converting visual text into a format the language model can process directly.

For purely text-based inputs or when no OCR engine is supplied, the extraction process will rely solely on the language model's text understanding capabilities.