sammeeey / ocr

Python script that converts PDFs into .txt files

OCR

built & tested using Python 3.11.2

Minimal version of Python script that finds PDF files in speficied directory, converts them into .txt files using Python Tesseract & saves final files in same directory as original PDF files.

Installation (on windows 10)

1. clone repo
2. enter repo directory: cd ocr
3. install Python tesseract
4. create virtual environment: py -m venv venv
5. activate virtual environment: venv\Scripts\activate.bat
6. update pip: py -m pip install --upgrade pip
7. install requirements: pip install -r requirements.txt
8. run program as described below (Usage)

Usage (on windows 10)

1. run py ocr.py
2. follow instructions & prompts of program

Quickstart

1. run py ocr.py
2. press Enter to use sample PDFs in ./samplePDFs subdirectory by default

What happens

ocr.py creates .txt files with content of all PDFs in given directory

• for each PDF in target directory (default: ./samplePDFs): pdf2image module used to convert PDFs in into images
  • for each image
    • pytesseract module used to convert text in image to string, then appends text to .txt file with name of original PDF & saves alongside original PDF
• process takes up to 10 minutes
• content of ./samplePDFs expected to look like ./samplePDFsResult eventually

Resources

• pytesseract installation
• guidance installation video
• sample application

sample PDFs

• https://www.africau.edu/images/default/sample.pdf
• https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf
• https://www.adobe.com/support/products/enterprise/knowledgecenter/media/c4611_sample_explain.pdf
• https://unec.edu.az/application/uploads/2014/12/pdf-sample.pdf
• https://file-examples.com/index.php/sample-documents-download/sample-pdf-download/
• https://www.orimi.com/pdf-test.pdf

Limitations / Known Issues

• potentially inaccurate - depending on quality, structure & content of input PDFs (images, charts, ...)

Potential Improvements

• adjust OCR settings to real world input PDFs (to achieve best results for expected input)
• create REST API to get share results with clients (long-term?)
  • authentication & encryption (e.g. using JSON Web Token)