Scan Extractor

This repo documents the process to extract strucutred data from images of textual data.

Step 1

Convert HEIC to JPG images one_two.sh

If your images are already in JPG format. You can skip this step. If your images are not in HEIC, you may be able to use ImageMagick's Convert program.

Step 2

Straighten, Dewarp, remove paper, and convert to Black & White I reccommend using Scan Tailor

If your images are really good. You might be able to use a vairation on two_three.sh which uses textcleaner by Fred Weinhaus

Step 3

Use Tesseract OCR to extract text from the images. three_four.sh

If you know your input has a limited character set, I have found that using

tessedit_char_whitelist eliminates post processing work needed.

Step 4

Post process the OCR output and generate a CSV. process.py