Confused about docling and OCR

[catmanjan]

I read the docling site and it mentions OCR - in the sense that you can pick your OCR model to use

does this mean docling doesnt come with an OCR engine? If it does, which ones does it come with and how do you trigger an OCR of an image for example?

[dosubot[bot]]

Hi @catmanjan! I'm Dosu and I'm helping the IBM Docling team.

Docling does come with built-in OCR engines. You don't need to bring your own. The bundled options are:

• EasyOCR — the default OCR engine for PDF/image processing [1]
• Tesseract (via CLI) [2]
• RapidOCR — a lightweight, CPU-friendly alternative [3]

There's also a KServe v2 option for remote OCR inference servers, and plugin support for external engines like SuryaOCR [4].

How to trigger OCR on an image or scanned PDF:

from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.datamodel.pipeline_options import PdfPipelineOptions, EasyOcrOptions
from docling.datamodel.base_models import InputFormat

pipeline_options = PdfPipelineOptions(
    do_ocr=True,
    ocr_options=EasyOcrOptions(lang=["en"])
)

converter = DocumentConverter(
    format_options={InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options)}
)

result = converter.convert("your_file.pdf")  # also works with image files
print(result.document.export_to_markdown())

OCR is enabled by setting do_ocr=True in the pipeline options. Docling automatically determines which pages need OCR (scanned/bitmap) versus direct text extraction (digital) [3]. For images, the StandardPdfPipeline is also used since Docling treats IMAGE format with the same pipeline as PDF [1].

To switch engines, just swap EasyOcrOptions for TesseractCliOcrOptions or RapidOcrOptions.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.

[catmanjan]

Can a human verify what is said above? So IBM docling has easyocr as a default dependency? Who is easyocr and how are they related to IBM?