description

This jupyter notebook script does the following:

1. preprocesses PDFs for OCR (i.e., deskew, auto-rotate, de-background, clean using ocrmypdf and unpaper),
2. OCRs the PDFs (using ocrmypdf/tesseract 4.1),
3. outputs the following files for every PDF in the directory except PDFs with extension .processed.pdf.
   • OCRed PDF/A with processing applyed (filename.processed.pdf)
   • text file with the pre-existing OCR of the original PDF if it exists (filename.original.txt)
   • the tesseract-generated OCR in a text file (filename.ocr.txt) for every PDF in the directory.
4. The script leaves the original PDFs unchanged.

usage

1. Paste OCR_PDFs.ipynb script in same directory with PDFs to be OCRed.
2. Run script by typing jupyter notebook OCR_PDFs.ipynb in terminal. (You will need jupyter, ocrmypdf, and tesseract installed.)
3. Run all code blocks in notebook.

nota bene

• Script will process all PDFs in the same directory, but will not recursively process PDFs in subfolders.
• Script uses shell commands (via the jupyter "!" command) alongside python 3 so this code will not work outside of a jupyter notebook.
• Script ignores PDFs named [filename].processed.pdf. If keep_processed_PDFs = False, it will overwrite all PDFs ending in processed.pdf.