PDF of image: how Camelot can help (or not)?

[piegu]

Hi,

firstly, I would like to thank all creators of Camelot (cc @vinayak-mehta ). It works very well with native digital PDF. :-)

Now, I would like to point to image PDF issue. Until now, I think that Camelot does not help.

Clearly, there are other models (like Table Transformer) to deal with image PDF that the result is not at the Camelot quality level.

"One day", Camelot will work on image PDF as well or never?
If never, what are the best tools/services/models for that?

Thank you.

[RazerP4antom]

Hi piegu,
How were you able to run camelot as when I am doing it I am getting a DeprecationError - : PdfFileReader is deprecated and was removed in PyPDF2 3.0.0. Use PdfReader instead.

[piegu]

Hi. Here my setup code:

!pip install ghostscript
!pip install camelot-py[cv]
!pip install -U "PyPDF2<3.0"
!apt install ghostscript python3-tk

[lipsa7]

Hi @piegu, can you please help me with an error?

I ran the code tables.export('foo.xlsx', f='excel'), but it is throwing me an error:
TypeError: NDFrame.to_excel() got an unexpected keyword argument 'encoding'

Can you please with this?

[piegu]

Hi @lipsa7,

Please, put your notebook on Colab and share the link here.
I will check the code.
Thank you.

[lipsa7]

Hi @piegu, it works on colab perfectly, just not on my machine
https://colab.research.google.com/drive/1Jy7w5fSNEie-ON7ZOIwGDxrHedTW9yuL#scrollTo=6mhGcsDX22L_

[bosd]

Closing — addressed via the docs path rather than parser changes.

Camelot still works only on text-based PDFs (the parser hot paths consume layout objects from playa). For image-based / scanned PDFs the recommended path is to pre-process with ocrmypdf, which adds a hidden text layer the rest of camelot can use unchanged:

```bash
$ ocrmypdf scanned.pdf scanned-ocr.pdf
```

```python

tables = camelot.read_pdf('scanned-ocr.pdf')
```

There's a full recipe with the recommended ocrmypdf flags in docs/user/advanced.rst ("Working with image-based / scanned PDFs"), added in #743. The FAQ entry in #748 also covers the "does Camelot work with image-based PDFs?" question with the same answer.

Native OCR in Camelot itself isn't planned — the in-tree OCR work in #209 is paused. The ocrmypdf-then-camelot pipeline handles ~all current use cases and keeps the parser implementation focused on text extraction.