ds4sd/DocLayNet · original width and height

Dec 12, 2023

Hello,

Here are the features:

{'image_id': Value(dtype='int64', id=None),
 'image': Image(decode=True, id=None),
 'width': Value(dtype='int32', id=None),
 'height': Value(dtype='int32', id=None),
 'doc_category': Value(dtype='string', id=None),
 'collection': Value(dtype='string', id=None),
 'doc_name': Value(dtype='string', id=None),
 'page_no': Value(dtype='int64', id=None),
 'objects': [{'category_id': ClassLabel(names=['Caption', 'Footnote', 'Formula', 'List-item', 'Page-footer', 'Page-header', 'Picture', 'Section-header', 'Table', 'Text', 'Title'], id=None),
   'image_id': Value(dtype='string', id=None),
   'id': Value(dtype='int64', id=None),
   'area': Value(dtype='int64', id=None),
   'bbox': Sequence(feature=Value(dtype='float32', id=None), length=4, id=None),
   'segmentation': [[Value(dtype='float32', id=None)]],
   'iscrowd': Value(dtype='bool', id=None),
   'precedence': Value(dtype='int32', id=None)}]}

Width and height are always 1025, which seems to correspond to the original dataset's "coco_width" and "coco_height".

However, I'm looking for the PDF pages' original widths and heights. Where is that included in this dataset?

asnassar

Deep Search org Feb 8

Hello, we have another format of the dataset here: https://huggingface.co/datasets/ds4sd/DocLayNet-v1.1, that includes in the metadata column the original dimensions.

asnassar changed discussion status to closed Feb 8

opyate

Feb 8

Thank you, I can now see "original_height": x, "original_width": y in the metadata.