PDF Image OCR

A Python Program extracting Images from a pdf-file and extracting the text using Optical Character Recognition. Afterwards, you may feed an LLM running locally via Ollama with the data.

Requirements

• The Project was made for Windows. If your OS is another like macOS or Linux, you need to adjust some paths because of \\.
• Python Version 3.11

pip install -r requirements.txt

• Install Tesseract and copy the path
• Install Ollama
• in console (example model: llama3.2):

ollama pull <YourModel>

Var.py

This is where the Paths are saved, change these according to your System

• tesseract_path: The Path to your tesseract.exe file (Windows Default: C:\\Program Files\\Tesseract-OCR\\tesseract.exe)
• input_path: Path to your PDF-File (example: "example_folder\\example.pdf")
• output_path: The extracted images and Text will go here (example: ".\\out\\")
• lang: language of the PDF (example: "deu+eng" for German and English)
• model: the ollama model you want to use (example: llama3.2 for English, cas/discolm-mfto-german for German)

Extras

• If you already have got images:
  • place them in your output_path folder
  • comment out the method call image_extraction() (to: #image_extraction())
  • delete the + folder in the line saying input_folder = output_path + folder
• If you only want to extract the images and do not want to use OCR/AI
  • comment out the method call ocr(input_folder, out_txt, lang) (to: #ocr(input_folder, out_txt, lang))
  • comment out the method call feedllm() (to: #feedllm())
• If you only want to use the AI with a txt file you have:
  • comment out the method call image_extraction() (to: #image_extraction())
  • delete the + folder in the line saying input_folder = output_path + folder
  • comment out the method call ocr(input_folder, out_txt, lang) (to: #ocr(input_folder, out_txt, lang))
  • place the .txt file in your output_path folder and rename it to out.txt