Skip to content

feat: convert Docling JSON inputs to image streams in FileDatasetBuilder #184

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
cau-git merged 2 commits into from
Dec 5, 2025
Merged

feat: convert Docling JSON inputs to image streams in FileDatasetBuilder #184

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
75bb690
feat: convert Docling JSON inputs to image streamsin FileDatasetBuilder
cau-git Dec 4, 2025
123adcc
Ensure usage of ImageFormatOptions where necessary
cau-git Dec 5, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 14 additions & 6 deletions docling_eval/cli/main.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@
from docling.datamodel.vlm_model_specs import (
SMOLDOCLING_TRANSFORMERS as smoldocling_vlm_conversion_options,
)
from docling.document_converter import FormatOption, PdfFormatOption
from docling.document_converter import FormatOption, ImageFormatOption, PdfFormatOption
from docling.models.factories import get_ocr_factory
from docling.pipeline.vlm_pipeline import VlmPipeline
from PyPDF2 import PdfReader, PdfWriter
Expand Down Expand Up @@ -414,7 +414,7 @@ def get_prediction_provider(
return DoclingPredictionProvider(
format_options={
InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options),
InputFormat.IMAGE: PdfFormatOption(pipeline_options=pipeline_options),
InputFormat.IMAGE: ImageFormatOption(pipeline_options=pipeline_options),
},
do_visualization=do_visualization,
ignore_missing_predictions=True,
Expand Down Expand Up @@ -444,7 +444,7 @@ def get_prediction_provider(
return DoclingPredictionProvider(
format_options={
InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options),
InputFormat.IMAGE: PdfFormatOption(pipeline_options=pipeline_options),
InputFormat.IMAGE: ImageFormatOption(pipeline_options=pipeline_options),
},
do_visualization=do_visualization,
ignore_missing_predictions=True,
Expand Down Expand Up @@ -493,7 +493,7 @@ def get_prediction_provider(
return DoclingPredictionProvider(
format_options={
InputFormat.PDF: PdfFormatOption(pipeline_options=pdf_pipeline_options),
InputFormat.IMAGE: PdfFormatOption(
InputFormat.IMAGE: ImageFormatOption(
pipeline_options=ocr_pipeline_options
),
},
Expand Down Expand Up @@ -528,10 +528,14 @@ def get_prediction_provider(
pipeline_cls=VlmPipeline, pipeline_options=pipeline_options
)

image_format_option = ImageFormatOption(
pipeline_cls=VlmPipeline, pipeline_options=pipeline_options
)

return DoclingPredictionProvider(
format_options={
InputFormat.PDF: pdf_format_option,
InputFormat.IMAGE: pdf_format_option,
InputFormat.IMAGE: image_format_option,
},
do_visualization=do_visualization,
ignore_missing_predictions=True,
Expand Down Expand Up @@ -575,10 +579,14 @@ def get_prediction_provider(
pipeline_cls=VlmPipeline, pipeline_options=pipeline_options
)

image_format_option = ImageFormatOption(
pipeline_cls=VlmPipeline, pipeline_options=pipeline_options
)

return DoclingPredictionProvider(
format_options={
InputFormat.PDF: pdf_format_option,
InputFormat.IMAGE: pdf_format_option,
InputFormat.IMAGE: image_format_option,
},
do_visualization=do_visualization,
ignore_missing_predictions=True,
Expand Down
45 changes: 39 additions & 6 deletions docling_eval/dataset_builders/file_dataset_builder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,11 +205,44 @@ def iterate(self) -> Iterable[DatasetRecord]:
page_images_column=BenchMarkColumns.GROUNDTRUTH_PAGE_IMAGES.value,
)

# Get source as binary data
source_bytes = get_binary(filename)
source_stream = DocumentStream(
name=filename.name, stream=BytesIO(source_bytes)
)
# Prepare source binary: for JSON inputs prefer image streams when page images exist
source_bytes: bytes
source_stream: DocumentStream
effective_mime_type = mime_type

if mime_type == "application/json" and len(true_page_images) > 0:
images_rgb = [
img.convert("RGB") if img.mode != "RGB" else img
for img in true_page_images
]

if len(images_rgb) == 1:
buffer = BytesIO()
images_rgb[0].save(buffer, format="PNG")
source_bytes = buffer.getvalue()
source_stream = DocumentStream(
name=f"{filename.stem}.png", stream=BytesIO(source_bytes)
)
effective_mime_type = "image/png"
else:
buffer = BytesIO()
images_rgb[0].save(
buffer,
format="TIFF",
save_all=True,
append_images=images_rgb[1:],
compression="tiff_lzw",
)
source_bytes = buffer.getvalue()
source_stream = DocumentStream(
name=f"{filename.stem}.tiff", stream=BytesIO(source_bytes)
)
effective_mime_type = "image/tiff"
else:
source_bytes = get_binary(filename)
source_stream = DocumentStream(
name=filename.name, stream=BytesIO(source_bytes)
)

# Create dataset record
record = DatasetRecord(
Expand All @@ -219,7 +252,7 @@ def iterate(self) -> Iterable[DatasetRecord]:
ground_truth_pictures=true_pictures,
ground_truth_page_images=true_page_images,
original=source_stream,
mime_type=mime_type,
mime_type=effective_mime_type,
)

yield record
Loading