Identification of images in docx

[jkindahood]

Hello everyone,
im working on docx document type with docling.
My first test was very poor.
I have a docx file which is only a testfile and this contains one image.
The standard procedure doesnt even recognise the image in the file...
Standard is this:

from docling.document_converter import DocumentConverter

source = "data/example_small.docx"
converter = DocumentConverter()
result = converter.convert(source)
for item in result.document:
print(item)

The example file is contained here!
Am im wrong with the code or is this a bug?

Greetings
example_small.docx
word_sample.docx

[PeterStaar-IBM]

@jkindahood Thanks for the feedback! Let us look into it and come back to you.

I dont see any handling of pictures in the msword-backend. If we need to add pictures (which we obviously need to do), we need to update handle_elements method.

[jkindahood]

can you describe me what to do?
maybe i can implement that.

[PeterStaar-IBM]

Yes, absolutely, if you follow the link, you see that we are have no add_picture method yet (as in the html version: https://github.com/DS4SD/docling/blob/main/docling/backend/html_backend.py#L429)

[jkindahood]

Thank you for your answer @PeterStaar-IBM.
I think this enhancement is stronly connected to this pull request:
#259
because im interested in the description of a picture, embedded into the text on the rigth position.
In a first step would you say we should read the images in wordbackend like in the pdf backend?
And in a second step we add the option to describe the image and add the description to the returned text?
Greetings

[maxmnemonic]

@jkindahood, small update, I'm working on PR to resolve image identification in DOCX: #330

[jkindahood]

@maxmnemonic all fine, i try to understand how your codebase works and still want to contribute.
Tell me if i can do some thing to help you.