/Python_OCR

Python OCR Implementation for a school project

Primary LanguagePythonOtherNOASSERTION

Py_OCR

An OCR using Python for School Assessed Coursework in Applied Computing

Demo

Installation/Documentation/Features

https://0x00sec.gitbook.io/py_ocr-documentation/installation/py_ocr-installaion

Py_OCR Only works with python3

Usage/Examples

python main.py --pdf PDF_Examples/testpdf2.pdf --denoise --gray --adapt

GUI Version

python ui/maingui.py

Best Results so far

python main.py --pdf "images\PDF_Examples\Gordon House [M3202]-Service-07-06-2022.pdf"  -b --psm 11

Authors

Lessons Learned

  • DO NOT build a GUI with only python, Django exists
  • Teamwork is important, it kept us going :D
  • Please for the love of god document the important bits and seperate everything into files
  • Watch out for circular imports :(

In general, we learnt loads about new libaries, inbuilt python standard libaries and dipped our toes into the wacky world of AI, Image Recognition and gained an appreciation for its complexity.

Release History/Bug Fixes

  • 1.0.0 - Yes!!! First release!!
    • Output image opens up in a new window, but all the other UI features work! Release #1!!
  • 0.3.1
    • I/O placeholders are working in GUI, Input placeholder also gets resized and updated. PDF's have not been tested.
  • 0.3.0
    • Switched from PyQt to Tkinter, eaisier to implement, GUI is ready to be hooked up to backend
  • 0.2.5
    • Linking the UI to maingui in progress, the buttons work but the displays don't for now. Work in Progess.
  • 0.2.1
    • main.gui file has been created to handle all the logic of the GUI file
  • 0.2.0
    • Tried implementing easyocr to work with pytesseract, their libraries clash and until a fix can be found, all of that code has been commented.
  • 0.1.9
    • Significant improvement in text detection by using https://tesseract-ocr.github.io/tessdoc/Data-Files tessdata-best instead of the default testdata training model along with custom oem, psm and adaptive thresholding options. tessdata-best will need to be manually installed in your local TESSERACT_OCR folder (and renamed to "tessdata") in Program Files until we implement a better solution
  • 0.1.8
    • Optimized the source code (about twice as fast)
  • 0.1.7
    • Adaptive thresholding command fixed | The config in image_to_text removed, it was causing a buttload of issues with no apparent solutions | Table extraction is work in progress
  • 0.1.6
    • Multipage pdfs can now be converted, converted pdfs are stored in the "frompdf" folder
  • 0.1.5
    • Pdfs can now be converted to images for text recognition (though it is very inaccurate)
  • 0.1.1
    • Fixed a major issue where the original image was being taken in for text conversion, and not the modified version (I'm dumb)
  • 0.1.0 - Initial closed-release with command-line features
    • Arguments can now be passed through the command-line to modify and customise the result as needed
  • 0.0.5
    • Word detection is now live :D
  • 0.0.4
    • Boxes around detected characters with prediction added
  • 0.0.3
    • image_to_string working, a bit inaccurate but fixable
  • 0.0.2
    • Image correction working, new requirements
  • 0.0.1
    • Base functionality (Work in progress)