Burmese-Table-OCR

💻 Extract text from tables of images. Use OpenCV to detect margin lines and PyTesseract to detect Burmese text. ⌨️

🏷️ To-Do List

• Upload to Github

• Folder Input

  • folder_input.py - current output as CSV append , so need to delete or find a way to overwrite to apply new folder. Then another possible work is directly upload into Google Sheets.
  • In output csv , there's uncessary characters(like \n symbol), need to find a way to remove these.(possible with pandas then overwrite)

• Link with Google APIs

• Google Sheets Helping Guide

• Fix not to overwrite for appending CSVs

• Always opening BW image for each page

  • I think I can fix with by changing waitkey() & destoryAllWindows functions. Solution - change waitkey(0) into waitkey(10) then add destory.

• Correct Horizontal , Vertical & Intersection of table

  • FindContour?
  • Adjusting parameters

• Pandas - Removing non-Unicode characters

  • Regex characters
  • Appending rows by rows
  • Even or Odd numbers of Array
    1. Dictionary into Dataframe
    2. 1D Array into Pandas.Dataframe series
    3. Combine multiple sereis as one Dataframe

• Accuracy Test

• Google Vision API

• Web Version

Error Notes 🔥 💦

• Unicode CSV encoding problem - when I try to export csv into google sheets , the font wasn't correct when using with Gspread 'import_csv' function.

  • Solution -> open("angel.csv", "r").read().encode("utf8")

• Nov 27,2020

  • A lot of errors also today. I didn't note down everything but the solved tasks that I remember is
    • Adjusting threshold & minLinLength values to detect the table correctly (it's the most important thing)
    • Append the dictionary according to filenames
    • Generate CSV - row by row
  • Overall result is satisfied.
  • My code is full of comments & editions. Noone won't be able to understand at the first look.:satisfied:
    • I need to write a blog about this project and also record an explanation video.

📚 Ref

1. Main Reference Guide
2. Burmese Tesseract Project