Extracting data from images using vision language models

Use VLM-enhanced ICR when you need higher extraction accuracy on complex documents.

Common use cases include:

• Financial documents with complex tables
• Invoices with varied layouts
• Medical records with specialized terminology
• Legal documents with strict structure requirements
• Multi-language document analysis

VLM-enhanced mode combines ICR layout analysis with language-model reasoning to improve classification and structure detection.

How Nutrient helps

Nutrient .NET SDK handles VLM-enhanced configuration, model orchestration, and JSON output generation.

The SDK handles:

• Hybrid mode configuration for ICR + VLM processing
• Model loading and capability coordination
• Semantic classification and confidence scoring internals
• Complex layout analysis implementation details

Prerequisites

Before running this sample, make sure your VLM setup is ready.

VLM-enhanced mode requires an external VLM endpoint. The SDK does not automatically provision or start a VLM service for you.

• Configure a reachable VLM endpoint in your environment.
• Set the ApiEndpoint and Model properties in custom VLM API settings.
• By default, the SDK may assume:
  • API endpoint: http://localhost:1234/v1
  • Model: qwen/qwen3-vl-8b
• For clarity and reliability, explicitly set both the API endpoint and the model in your configuration.
• Example with LM Studio(opens in a new tab):
  • Run LM Studio in server mode.
  • Load a compatible vision model such as Qwen3-VL (4B, 8B, or 23B depending on your hardware).
  • API endpoint: http://127.0.0.1:1234/v1
  • Model: qwen/qwen3-vl-4b
• Ensure the endpoint is running before calling ExtractContent() in VLM-enhanced mode.

If no VLM endpoint is available, VLM-enhanced extraction can fail at runtime.

Complete implementation

This example extracts structured JSON using VisionEngine.VlmEnhancedIcr:

using Nutrient;

Loading and processing the image

Open the image with a using statement(opens in a new tab) so the handle is released when the block ends:

try
{
    using Document document = Document.Open("input.png");

Configuring VLM-enhanced mode

Set the vision engine to VisionEngine.VlmEnhancedIcr.

This mode improves:

• Table boundary detection
• Semantic element classification
• Reading order in complex layouts
• Understanding across document variations

    document.Settings.VisionSettings.Engine = VisionEngine.VlmEnhancedIcr;

Creating a vision instance

Create a vision instance bound to the opened document with Vision.Set(document):

    var vision = Vision.Set(document);

Extracting structured content

Call ExtractContent() to run the VLM-enhanced pipeline.

In this mode, the pipeline performs:

• Initial ICR layout detection
• VLM-based semantic refinement
• Confidence scoring
• JSON generation with structure and coordinates

    string contentJson = vision.ExtractContent();

Write the JSON result to a file for downstream processing.

Use this output for indexing, validation, storage, or custom analysis:

    File.WriteAllText("output.json", contentJson);
}
catch (NutrientException e)
{
    Console.Error.WriteLine($"Error: {e.Message}");
    Environment.Exit(1);
}

Understanding the output

ExtractContent() returns structured JSON with layout and semantic metadata.

VLM-enhanced output includes:

• Document elements — Paragraphs, headings, tables, figures, equations, detected barcodes, and form-related regions
• Barcode data — Decoded barcode values with symbology information for supported 1D and 2D barcode types
• Bounding boxes — Pixel coordinates with improved boundary accuracy
• Hierarchical relationships — Parent-child structure across sections and blocks
• Element classification — Semantic types with confidence scores
• Reading order — Sequence for complex layouts and multicolumn content
• Semantic metadata — Additional attributes used in downstream processing

Key output fields

The following are the most commonly included fields in VLM JSON output:

• text — Extracted text for the element.
• words — Per-word OCR/extraction results.
• bounds — Bounding box coordinates for the element or word.
• confidence — Confidence score for the element or word.
• readingOrder — Sequence in which elements should be read.
• id — Unique identifier for the extracted element.
• pageNumber — Source page number.
• type / role — Semantic type of the extracted block, such as text, heading, table, image, or barcode.

When an element contains only one word, element-level and word-level bounds/confidence can appear identical.

Confidence fields in VLM output

VLM output can contain two distinct confidence signals:

1. confidence (or classificationConfidence) — Zone classification confidence
   • Definition: How confident the model is in semantic classification (for example, text, heading, table, image), heading level detection, and language detection.
   • Scale: 0.0 to 1.0 (float).
   • Interpretation:
     • 0.0 = no confidence (often treated as unknown classification)
     • 1.0 = maximum confidence
   • Use: Decide whether to trust semantic zone labels in downstream logic.
2. textConfidence — Text extraction confidence
   • Definition: How confident the model is in the extracted text quality for a zone.
   • Scale: Categorical values: high, medium, low (not numeric).
   • Interpretation:
     • high = strong confidence in extracted text
     • medium = moderate confidence
     • low = uncertain text quality
   • Use: Prioritize review, fallback, or fusion strategies for lower-confidence text.

Use this JSON for form extraction, contract analysis, invoice parsing, and other high-accuracy workflows.

Error handling

Vision API throws a NutrientException when extraction fails.

Common failure scenarios include:

• The image file can’t be read due to path or permission issues
• Image data is corrupted or unsupported
• Required models are missing or inaccessible
• Available memory is insufficient for VLM-enhanced processing
• VLM enhancement fails due to connectivity or service issues when applicable
• Image format, resolution, or dimensions are unsupported

In production code:

• Catch NutrientException.
• Return a clear error message.
• Log failure details for debugging.
• Add fallback logic (for example, retry in ICR mode).

Conclusion

Use this workflow for VLM-enhanced extraction:

1. Open the image document with a using statement for automatic resource cleanup.
2. Set Engine to VisionEngine.VlmEnhancedIcr in the vision settings for enhanced accuracy.
3. VLM-enhanced mode combines local ICR AI models with vision language model capabilities for superior document analysis.
4. Create a vision instance with Vision.Set() to bind content extraction operations to the document.
5. Call ExtractContent() to invoke the VLM-enhanced processing pipeline.
6. The pipeline performs initial ICR layout analysis, applies VLM enhancement for semantic understanding, calculates confidence scores, and generates JSON output.
7. VLM enhancement improves table cell boundary detection, element classification accuracy, and reading order determination for complex layouts.
8. The method returns a JSON-formatted string containing document structure with elements, bounding boxes, hierarchical relationships, reading order, and confidence scores.
9. Write the JSON content to a file for intelligent form extraction, contract analysis, invoice processing, and legal document parsing.
10. Handle NutrientException failures for robust error recovery with fallback strategies like pure ICR mode.
11. VLM-enhanced mode is ideal for complex documents where extraction accuracy is the priority.

For related image extraction workflows, refer to the .NET SDK guides.

Download this ready-to-use sample package to explore VLM-enhanced extraction.