PDF to Markdown - Images Not Extracted, Showing <--image--> Placeholder

[harskuma]

I am trying to convert a PDF to Markdown using Docling, but instead of extracting the images, I am getting a <--image--> placeholder in the output.

Used pipeline_options.do_ocr = True to enable OCR.

[jaddison]

See the image_mode parameter to export_to_markdown. There are 3 options, one of which is REFERENCED.

class ImageRefMode(str, Enum):
    """ImageRefMode."""

    PLACEHOLDER = "placeholder"  # just a place-holder
    EMBEDDED = "embedded"  # embed the image as a base64
    REFERENCED = "referenced"  # reference the image via uri

[KartikB4B]

when you are exporting the markdown using the doc.export_to_markdown(),you can specify the image_mode.
Example ->

full_markdown = doc.export_to_markdown(
page_break_placeholder="",
image_mode="embedded"
)

[harskuma]

Hey! Thanks for the reference , i used this but i'm getting image as base64, i want the text from images,am i missing something?

[KartikB4B]

No . I am also looking for that.

[sreena-certaintiai]

Same issue goes here.

[itsyaboyksi]

Any luck yet?

[harskuma]

@itsyaboyksi @sreena-certaintiai @itsyaboyksi
Hey! I was just looking around and found a way to access the image content
Here's what worked for me (I used the sample PDF amt_handbook_sample.pdf from the Docling repo):

source = "amt_handbook_sample.pdf"

pipeline_options = PdfPipelineOptions()
pipeline_options.images_scale = 2
pipeline_options.generate_page_images = True

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

result = doc_converter.convert(source)
doc = result.document

for picture in doc.pictures:
    print(picture.caption_text(doc), " contains these elements:")
    for item, level in doc.iterate_items(root=picture, traverse_pictures=True):
        if isinstance(item, TextItem):
            print(item.text)
    print("\n")

This gives you access to the image blocks and any OCR-extracted text inside them.
For now, I'm replacing the placeholder in the Markdown with the output of this code using some custom logic.