bug/partition-pdf-with-infer_table_structure

[DeepKariaX]

Describe the bug
Giving (ValueError: max() arg is an empty sequence) error when using partition pdf. When i keep the infer_table_structure = True parameter it is giving me this error and after removing this parameter it is working perfectly.

File which received bug
unstructured_inference/models/tables.py", line 667, in fill_cells
table_rows_no = max({row for cell in cells for row in cell["row_nums"]})

Expected behavior
Even if we keep the infer_table_structure = True parameter it should be able to partition the pdf without any errors. (Maybe add error handling when receiving the none value)

[vav1lo]

we got the same issue too. is there any solution ?

[DeepKariaX]

@vav1lo Currently, I have changed to another reader. Also can you attach the pdf which you are testing coz mine is bit confidential to share and with a sample pdf it would be easy for them to diagnose the error.

[christinestraub]

Hi @vav1lo, Can you please attach the pdf that you are testing?

[hackpointt]

uber_10q_march_2022.pdf
same problem with this file

`import os
from unstructured.partition.pdf import partition_pdf
from unstructured.staging.base import elements_to_json

filename = "uber_10q_march_2022.pdf"

elements = partition_pdf(
filename=filename,
strategy="hi_res",
infer_table_structure=True,
model_name="yolox",
)`

[hackpointt]

@christinestraub

[vav1lo]

@christinestraub Here is the pdf that i am testing
1b4c03d6-f6f5-462d-8bd6-0b9e411bc33d.pdf

[Nidhi2497]

I am also getting the error while partitioning pdf , and the error is with particularly this argument infer_table_structure=True,

 9 import torch
 10 import transformers

---> 11 from cv2.typing import MatLike
12 from PIL.Image import Image
13 from transformers import DonutProcessor, VisionEncoderDecoderModel

ModuleNotFoundError: No module named 'cv2.typing'; 'cv2' is not a package

[vav1lo]

I am also getting the error while partitioning pdf , and the error is with particularly this argument infer_table_structure=True,

 9 import torch
 10 import transformers

---> 11 from cv2.typing import MatLike 12 from PIL.Image import Image 13 from transformers import DonutProcessor, VisionEncoderDecoderModel

ModuleNotFoundError: No module named 'cv2.typing'; 'cv2' is not a package

I think this has to do with the opencv installation

[nikklavzar]

we got the same issue too. is there any solution ?

This started happening to me when I upgraded from 0.12.6 to 0.14.6

[Nidhi2497]

I am also getting the error while partitioning pdf , and the error is with particularly this argument infer_table_structure=True,

 9 import torch
 10 import transformers

---> 11 from cv2.typing import MatLike 12 from PIL.Image import Image 13 from transformers import DonutProcessor, VisionEncoderDecoderModel
ModuleNotFoundError: No module named 'cv2.typing'; 'cv2' is not a package

I think this has to do with the opencv installation

i installed it as well, but what is being imported there needs to be changed actually

[christinestraub]

Hi @DeepKariaX, @vav1lo, @hackpointt, @Nidhi2497, @nikklavzar

Addressed on Unstructured-IO/unstructured-inference#359. You'll need to upgrade unstructured-inference to 0.7.36. I tested your code with the provided pdf documents and it worked as expected.

[christinestraub]

Closing this since it's assumed to be resolved, but feel free to reopen if you're still having this issue.

[DeepKariaX]

@christinestraub This is resolved, thanks !