AI GitHub
MinerU for extracting PDF layout and formulas
Convert complex documents to structured outputs for reading or downstream processing.
Key Takeaways
- MinerU combines layout analysis and content extraction to recover items such as equations, tables and text blocks.
- Review backend and model choices for the input language and hardware.
- Choose a paper with a displayed equation and a table that spans columns.
- A correct-looking export can silently lose subscripts or reorder columns.
How it works
MinerU combines layout analysis and content extraction to recover items such as equations, tables and text blocks. Its output can be inspected alongside the source page.
What you need
Review backend and model choices for the input language and hardware. Scanned documents and born-digital PDFs may exercise different paths through the extraction pipeline.
What to try first
Choose a paper with a displayed equation and a table that spans columns. Verify the extracted order and check the equation symbols before indexing the result.
Check an extracted equation against the typeset source and a simple numeric interpretation. A single lost minus sign can reverse a result.
Where to look for failures
A correct-looking export can silently lose subscripts or reorder columns. Confirm the pipeline and model requirements for your device and document language.
Related Franklin pages
Run the same scan through each approach and inspect the failed regions. An OCR engine and a full document-conversion pipeline can differ substantially in reading order and structural output.
- Convert PDFs and supported files into Markdown, JSON or other documented formats. Marker for converting documents to structured text
- Detect text and analyze layout, reading order and tables in documents. Surya for OCR and document layout
For how recognizing characters fits into the larger document processing workflow, see Franklin’s glossary. Optical character recognition
Project documentation
Try it for your workflow
Explore the project’s documentation for setup and usage, or save this page to a collection for later.
Read the project documentation ↗