Python Libraries to Extract Tables from PDFs

In this video we compare different packages and strategies for extracting tables from PDF documents in Python.

◾◾◾◾◾◾◾◾◾◾◾◾◾◾◾◾◾
📚 Programming Books & Merch 📚

💼 Services 💼

🌐 Social Media & Contact 🌐

Timestamps:
(0:00) Intro
(0:23) PDF Documents
(2:43) Camelot
(7:46) Tabula
(10:55) PDFPlumber
(17:16) LLMWhisperer
(23:32) PyPDF2
(26:40) Unstract

[bloody_albatross]

Great video! Some times tables are so dense so that the gap between columns is at places not more than the gap between words within a cell. Some tools have problems with that. Would have liked to see how these tools deal with that.

[djl]

I find Tabula (Java web app version) works best for my needs. I tried several Python-based PDF table extractors but the output was too unpredictable and/or inaccurate. Unfortunately Tabula's dependency on old (unsupported) Java versions makes it difficult to use on more recent Ubuntu releases. Coincidentally, just this morning I built a docker image that runs the Tabula Java web app on my Ubuntu 24.04 install -- once again Docker has proved to be a really useful tool!

[rohithreddy]

Just a suggestion : Please make a intro automation showing the result of your title so i can exactly know what i am getting into before watching a 30 min video Though the title itself is self explanatory here sometimes its not.

[icholakov]

nice video but your table examples are pretty simplistic...try a financial statement with three rows of cascading headers. while an invoice is a table...it hardly is a real representation of a table. ML based chips have been extracting data from invoices for 20 yrs now. my choice, after many attempts, was docling

[uwegenosdude]

Cool. Thanks a lot for your video.
Does llm whisperer upload my pdf to an external AI hoster to do this great job?

[marbacc]

Hello. Does this work with PDF's that have tables as images and not as proper tables?

[davidtindell]

Unfortunately, "privacy" is a major concern when extracting tables from personal or business PDF's !

[АнуарНаурызбаев-мщ]

Why not use chat gpt directly? In combination with pypdf it is possible to crop needed pages and send them to gpt. The LLMWhisperer overall not bad I think. Good work! Pls make video about enlargin vram of gpu!