generate_page_image from PPTX?

[tejaswib428]

I have been trying to extract pages as images from PPTs, but have failed so far. Does PDFPipelineOptions class for pipelines and PDFFormatOption class for Document Conversion work in case of PPTs?

I tried this way, I was able to extract text and tables but images were unsuccessful.

Then I tried to implement PPTXFormatOption class, but am doubtful which PipelineOptions class to use.

[dolfim-ibm]

Can you please post

1. the example code you are running
2. the docling version you are using docling --version

there have been some recent fixes which might have addressed this already.

[tejaswib428]

Hi, sure

The following is the code I was using:

`import time
import json
import pandas as pd
from pathlib import Path
from docling_core.types.doc.document import PageItem
from docling_core.types.doc import ImageRefMode, ImageRef, PictureItem, TableItem
from docling.datamodel.base_models import FigureElement, InputFormat, Table
from docling.document_converter import DocumentConverter, PdfFormatOption, PowerpointFormatOption
from docling.datamodel.pipeline_options import PdfPipelineOptions, TableFormerMode

def convert_document(ppt_path, output_dir):
"""
Converts a PPTX document to JSON format and extracts tables as DataFrames.

Args:
ppt_path (str): Path to the PPTX document.
output_dir_path (str): Path to the output directory.

Returns:
None
"""

# Set up pipeline options
pipeline_options = PdfPipelineOptions(
    do_table_structure=True,
    table_structure_options={"mode": TableFormerMode.ACCURATE}
)
pipeline_options.table_structure_options.do_cell_matching = False
pipeline_options.images_scale = 2.0
pipeline_options.generate_page_images = True

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

start_time = time.time()
conv_result = doc_converter.convert(ppt_path)
end_time = time.time() - start_time

logger.info(f"Document converted in {end_time:.2f} seconds.")

# Extracting tables as DataFrames
tables = conv_result.document.tables
doc_filename = conv_result.input.file.stem

table_dfs = {}  # Dictionary to hold DataFrames

image_folder = Path(f"{output_dir}/{doc_filename}_images")
image_folder.mkdir(parents = True, exist_ok = True)

for page_no, page in conv_result.document.pages.items():
    page_no = page.page_no
    page_image_filename = image_folder / f"{doc_filename}_{page_no}.png"
    with page_image_filename.open("wb") as fp:
        page.image.pil_image.save(fp, format="PNG")

if tables:
    logger.info(f"Available attributes in tables: {dir(tables)}")
    for table_ix, table in enumerate(tables):
        table_df: pd.DataFrame = table.export_to_dataframe()
        table_dfs[f"table_{table_ix}"] = table_df  # Store DataFrame in the dictionary
        logger.info(f"## Table {table_ix} extracted.")

    # Save the entire document as a JSON file
    output_dir = Path(output_dir_path)
    output_dir.mkdir(parents=True, exist_ok=True)

    output_path = Path(f"{output_dir}/{doc_filename}_parsed.json")
    with output_path.open("w", encoding="utf-8") as fp:
        fp.write(json.dumps(conv_result.document.export_to_dict()))

    

    # Return the dictionary of DataFrames
    return output_path, table_dfs, doc_filename
else:
    logger.info("No tables found in the document.")
    return None
    `

docling version:

Docling version: 2.7.0
Docling Core version: 2.4.1
Docling IBM Models version: 2.0.6
Docling Parse version: 2.1.2

[dolfim-ibm]

This code is saving the whole page images, which indeed are not supported in PPTX.

But you should still be able to save the images in the document with something like the following:

    for element, _level in conv_res.document.iterate_items():
        if isinstance(element, TableItem):
            table_counter += 1
            element_image_filename = (
                output_dir / f"{doc_filename}-table-{table_counter}.png"
            )
            with element_image_filename.open("wb") as fp:
                element.get_image(conv_res.document).save(fp, "PNG")

        if isinstance(element, PictureItem):
            picture_counter += 1
            element_image_filename = (
                output_dir / f"{doc_filename}-picture-{picture_counter}.png"
            )
            with element_image_filename.open("wb") as fp:
                element.get_image(conv_res.document).save(fp, "PNG")

The full example is here: https://ds4sd.github.io/docling/examples/export_figures/

[tejaswib428]

Saving pages as images from PPTX was my requirement. Thanks for the clarification.