Document type detection

Aspose.OCR for Python via .NET can automatically classify the content of an image or document page before running a specialized OCR workflow. This helps route mixed input batches to the most suitable recognition or post-processing pipeline.

Document type detection is available in two modes:

• AsposeOcr.detect_document_type() uses built-in rule-based logic combined with a neural model.
• AsposeOcr.detect_document_type_ai() uses an AI-powered LLM workflow and returns the model response as an AIResult.

Supported document types

The following document categories can be detected:

Document type	Description
DocType.UNKNOWN	Document type was not detected.
DocType.PICTURE	Photo, advertisement, billboard, packaging, or similar visual content.
DocType.HANDWRITTEN	Handwritten content.
DocType.BOOK	Book page or other primarily textual content.
DocType.FORMULA	Mathematical formula, equation, or expression.
DocType.TABLE	Table, form, or spreadsheet-like content.
DocType.PRESENTATION	Presentation slide.
DocType.SCIENTIFIC	Scientific article, report, paper, or technical publication.
DocType.INVOICE	Invoice, bill, receipt, or payment document.

Detecting document type

To detect document type, provide the collection of images to detect_document_type() method.

The method returns a list of DocTypeOutput objects. Each item contains:

Property	Description
source	Image path or URL when available. Empty for in-memory input.
page	Zero-based page number.
doc_type	Detected document category.
confidence	Confidence score from 0.0 to 1.0.

import aspose.ocr

api = aspose.ocr.AsposeOcr()

input_data = aspose.ocr.OcrInput(aspose.ocr.InputType.SINGLE_IMAGE)
input_data.add("invoice.png")

results = api.detect_document_type(input_data)

for result in results:
    print("Source:", result.source)
    print("Page:", result.page)
    print("Document type:", result.doc_type)
    print(f"Confidence: {result.confidence:.0%}")

Detecting document type with AI

Use detect_document_type_ai() to classify documents with an AI-powered LLM workflow.

The method returns a list of aspose.ocr.ai.AIResult objects. The result property contains the model response, usually as a JSON string with document type, confidence, and reasoning.

import aspose.ocr

api = aspose.ocr.AsposeOcr()

input_data = aspose.ocr.OcrInput(aspose.ocr.InputType.SINGLE_IMAGE)
input_data.add("invoice.png")

results = api.detect_document_type_ai(input_data)

for result in results:
    print("File:", result.file_name)
    print(result.result)

Detecting document types in a PDF

For multi-page documents, one result item is returned for each processed page.

import aspose.ocr

api = aspose.ocr.AsposeOcr()

input_data = aspose.ocr.OcrInput(aspose.ocr.InputType.PDF)
input_data.add("mixed-document.pdf", 0, 3)

results = api.detect_document_type(input_data)

for result in results:
    print(f"Page {result.page}: {result.doc_type} ({result.confidence:.0%})")