peetio / diard Goto Github PK

Hi, I have issues when using TableExtractor.

first, the bottom_y = rows[-1]["bbox"][3] in tables.py line 142 variable seems not to be used in any of the code, or i am wrong?

after I comment on that line, it works, and into the second one

second, why i always get empty result pandas.errors.EmptyDataError: No columns to parse from file? trying to lower the threshold but it is still the same.

and last thing, maybe you could check roi = image.crop((xmin, ymin, xmax, ymax)) in tables.py line 56, i get this error ValueError: Coordinate 'lower' is less than 'upper' and i try to change the values like this one roi = image.crop((xmax, ymax, xmin, ymin))

The layout detected table on pdf but when I enable table detection, it throws the following error.

Traceback (most recent call last):

  File "/var/folders/kt/7g6lg6l15s96drlgk_p123tc0000gp/T/ipykernel_81639/3626377987.py", line 2, in <module>
    doc.extract_layouts(visualize=True, segment_sections=True)

  File "/Users/rmehta/Diard/modules/document.py", line 361, in extract_layouts
    self.__set_extracted_data(b, img, page, j)

  File "/Users/rmehta/Diard/modules/document.py", line 262, in __set_extracted_data
    snippet, lang=self.lang, debug=False, threshold=0.7

  File "/Users/rmehta/Diard/modules/tables.py", line 420, in get_table_data
    objs = structure_table(objs, table_bbox)

  File "/Users/rmehta/Diard/modules/tables.py", line 142, in structure_table
    bottom_y = rows[-1]["bbox"][3]

IndexError: list index out of range

The pretrained model I used has been downloaded as given in instructions. pubtables1m_detection_detr_r18.pth

Traceback (most recent call last):
  File "main.py", line 8, in <module>
    from modules.tables import TableExtractor
  File "/*/Diard/modules/tables.py", line 18, in <module>
    import datasets.transforms as R
ModuleNotFoundError: No module named 'datasets.transforms'

An error occurs when running the code：#python main.py --overwrite --skip-failures

Thanks for the code. There's one question from my side that given all the bounding boxes for the image, how can we fine tune DiT on objection detection task?

Let's suppose I want to predict 2 different types of BBoxes on my images, how can I do that? What would be the changes in config file, code etc? I don't have masks nor do I want to predict. Just the bounding boxes.

I have asked a similar question here

Any blog, code, your code etc etc would be helpful.

thanks in advance :)